CN108648827A - Cardiovascular and cerebrovascular disease Risk Forecast Method and device - Google Patents

Abstract

A kind of cardiovascular and cerebrovascular disease Risk Forecast Method provided in an embodiment of the present invention and device, including：Obtain sample set；Sample in sample set is divided into the local cluster of preset quantity, according to preset first K values and first distance set, the first K values first of input sample are calculated adjacent to sample, so that it is determined that target part cluster, calculate the input sample in the cluster of the target part at a distance from sample, so that it is determined that the 2nd K values second of the input sample are adjacent to sample；The label for determining input sample, determine input sample whether be Patients with Cardiovascular/Cerebrovascular Diseases sample；Finally determine whether patient to be predicted is Patients with Cardiovascular/Cerebrovascular Diseases.The present embodiment is higher in view of Patients with Cardiovascular/Cerebrovascular Diseases characteristic similarity, avoids influence of the different sample datas to training prediction model.It is thus possible to improve predict patient to be predicted whether be Patients with Cardiovascular/Cerebrovascular Diseases accuracy rate.

Description

Risk prediction method and device for cardiovascular and cerebrovascular diseases

technical field

The invention relates to the field of predictive analysis, in particular to a method and device for predicting the risk of cardiovascular and cerebrovascular diseases.

Background technique

With the increasing life pressure and mental pressure of people, the incidence of cardiovascular and cerebrovascular diseases has increased year by year, seriously affecting the health of residents. Medical practice shows that if patients with cardiovascular and cerebrovascular diseases can be diagnosed at an early stage, it will be of great help to the intervention and treatment of cardiovascular and cerebrovascular diseases.

The existing technology uses data mining technology to mine the characteristics of the case data of cardiovascular and cerebrovascular diseases, and forms a training set from the physical examination feature data and return visit data of all patients, and uses decision tree, logistic regression and artificial neural network algorithms to train predictive model. Then, the physical examination data of the patient to be predicted is used as an input sample, which is input into the trained prediction model, and whether the patient to be predicted is a cardiovascular and cerebrovascular disease patient is output.

Taking the artificial neural network algorithm training prediction model as an example, in the process of using the artificial neural network algorithm to train the prediction model, since the input samples of the neural network include samples of patients with non-cardiovascular and cerebrovascular diseases and samples of patients with cardiovascular and cerebrovascular diseases, rather than samples of patients with cardiovascular and cerebrovascular diseases. There is a large gap between the characteristic data of the samples of patients with diseases and the samples of patients with cardiovascular and cerebrovascular diseases. Therefore, if all the samples in the training set are used as the input of the input layer, the error function of the output layer of the neural network is relatively large. Due to the influence of different sample data, the weights and thresholds of each layer of the neural network are adjusted according to the error function, and the trained prediction model is not accurate. Therefore, the accuracy rate of using the artificial neural network algorithm to train the prediction model to predict whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases is not high.

Contents of the invention

The purpose of the embodiments of the present invention is to provide a cardiovascular and cerebrovascular disease risk prediction method and device, so as to improve the accuracy of predicting whether a patient is a cardiovascular and cerebrovascular disease patient. The specific technical scheme is as follows:

In the first aspect, the embodiment of the present invention provides a method for predicting the risk of cardiovascular and cerebrovascular diseases, including:

Obtain a sample set; the sample set is determined according to multiple samples of the patient medical database set with labels; one sample includes: the patient's number, characteristics and characteristic data; the labels include: the first label and the second label ; The first label identifies samples from patients with cardiovascular and cerebrovascular diseases; the second label identifies samples from patients with non-cardio-cerebrovascular diseases;

Obtain an input sample; the input sample is formed by merging the medical health examination data and medical visit data of the patient to be predicted;

Use the cosine-large interval nearest neighbor COS-LMNN algorithm to carry out metric learning to obtain the global metric matrix of the sample set;

Use a preset clustering algorithm to divide the samples in the sample set into a preset number of local clusters;

According to the global metric matrix, using a cosine similarity algorithm to calculate the distance between the input sample and the samples in the sample set to form a first distance set;

According to the preset first K value and the first distance set, use the k-nearest neighbor algorithm to calculate the first K-value first neighboring samples of the input sample;

determining the local cluster in which the first neighboring sample is located;

In the local cluster where the first adjacent samples are located, select a local cluster whose number of first adjacent samples exceeds a first preset threshold as a target local cluster;

classifying the input samples into the target local clusters;

The local metric matrix of the target local cluster learned according to the COS-LMNN algorithm is calculated using a cosine similarity algorithm, and the distance between the input sample and the samples in the target local cluster forms a second distance set;

In the target local cluster, according to the preset second K value and the second distance set, use the k-nearest neighbor algorithm to determine the second K-value second neighboring samples of the input sample;

counting the number of the first label and the number of the second label of the second adjacent sample;

If the ratio of the number of the first label to the number of the second label exceeds the preset label threshold, the first label is used as the label of the input sample, otherwise the second label is used as the label of the input sample;

determining whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases according to the label of the input sample;

If the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases, it is determined that the patient to be predicted in the input sample is a patient with cardiovascular and cerebrovascular diseases;

If the input sample is not a sample of a patient with cardiovascular and cerebrovascular diseases, it is determined that the patient to be predicted in the input sample is not a patient with cardiovascular and cerebrovascular diseases.

Optionally, after the step of determining that the patient to be predicted in the input sample is a patient with cardiovascular and cerebrovascular diseases, the method further includes:

Determine whether the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient according to the patient's health follow-up data;

If the patient to be predicted is a patient with high-risk cardiovascular and cerebrovascular diseases, a suggestion for hospitalization is made to the patient to be predicted;

If the patient to be predicted is not a high-risk cardiovascular and cerebrovascular disease patient, a suggestion to increase the frequency of physical examination is made to the patient to be predicted;

After the step of determining that the patient to be predicted in the input sample is not a patient with cardiovascular and cerebrovascular diseases, the method also includes:

Determine whether the patient to be predicted is a healthy user according to the patient's health follow-up data;

If the patient to be predicted is a healthy user, make a suggestion for the normal patient to maintain a normal frequency of physical examination;

If the patient to be predicted is not a healthy user, then mark the patient to be predicted as a missed diagnosis patient, and add the feature data of the missed diagnosis patient to the patient medical database set;

Among them, the missed diagnosis patients were patients with cardiovascular and cerebrovascular diseases.

Optionally, the first label identifies samples of patients with cardiovascular and cerebrovascular diseases, including:

Determine the identification information of patients with cardiovascular and cerebrovascular diseases based on the collected patient health follow-up data;

The health follow-up data of the patient includes: the patient's number, characteristics, characteristic data, and confirmed symptoms; the identification information includes: confirmed symptoms, characteristics and characteristic data corresponding to the confirmed symptoms;

According to the identification information of patients with cardiovascular and cerebrovascular diseases, determine the samples of patients with cardiovascular and cerebrovascular diseases in the medical database set;

Set the first label on the samples of patients with cardiovascular and cerebrovascular diseases;

The second label identifies non-cardiovascular and cerebrovascular disease patient samples, including:

Set the second label on other samples except the samples of patients with cardiovascular and cerebrovascular diseases.

Optionally, obtain a sample set, including:

According to the multiple samples of the patient medical database set with labels, the samples whose missing values are greater than the first threshold are deleted as samples;

The sample missing value is: the ratio of the number of missing features in a sample to the total number of features in the sample;

Search in multiple samples after deletion processing, and perform feature deletion processing for features whose feature missing value is greater than the second threshold;

The feature missing value is: in the same feature of multiple samples, the ratio of the number of features lacking feature data to the total number of the same feature;

Find the feature of missing feature data in multiple samples after feature deletion processing, as the first feature;

Using a multiple filling method to fill in missing values for the feature data missing from the first feature;

According to the data type, classify the characteristic data of the plurality of samples after the missing value is filled, and obtain a classification result;

Wherein, the classification result includes: discrete feature data and continuous feature data;

According to the classification result, the discrete feature data and the continuous feature data are processed corresponding to the data type;

Adding the processed feature data corresponding to the discrete feature data and the continuous feature data to the patient medical database set as the first database set;

Wherein, the discrete feature data and the continuous feature data are processed corresponding to the data type, including: performing one-hot encoding on the discrete feature data; standardizing the continuous feature data using the normalized z-score method;

Using the undersampling and SMOTE algorithm to perform unbalanced processing on the samples of the first database set to obtain the second database set;

Using the variance analysis method to calculate the variance of the same feature data in the second database set, and delete the feature data whose variance value of the feature data is less than a preset variance threshold;

Using a relief algorithm to calculate the weight of each feature data after the feature data whose variance value is less than a preset variance threshold is deleted;

According to the weight of the feature data and the score value corresponding to the weight of the feature data, the feature data and the corresponding feature whose score value is less than the preset score threshold are deleted to obtain the fourth database set;

According to the fourth database set, a forward selection method is used to determine a sample set.

Optionally, according to the global metric matrix, the distance between the input sample and the samples in the sample set is calculated using a cosine similarity algorithm to form a first distance set, including:

According to the global metric matrix, using a cosine similarity algorithm formula to calculate the distance between the input sample and the samples in the sample set to form a first distance set;

Wherein, the cosine similarity algorithm formula is:

The first distance set D1 includes: {D( _xi ,x ₁ ), D( _xi ,x ₂ ), D( _xi ,x ₃ ),...,D( _xi ,x _j )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x _i ; the sample set is X; the global metric matrix is A; M= ^AT A; j represents the sample number in the sample set; x _j represents the sample The j-th sample in the set; i and j take positive integers; D( _xi , x _j ) represents the distance between the input sample x _i and the j-th sample in the X set under the global metric matrix; A( _xi , x _j ) Represents the distance between x _i and x _j after A matrix transformation.

Optionally, the local metric matrix of the target local cluster learned according to the COS-LMNN algorithm uses a cosine similarity algorithm to calculate the distance between the input sample and the samples in the target local cluster to form a second distance set ,include:

According to the local metric matrix of the target local cluster learned by the COS-LMNN algorithm, using the cosine similarity algorithm formula to calculate the distance between the input sample and the samples in the sample set to form a second distance set;

Wherein, the cosine similarity algorithm formula is:

The second distance set D2 includes: {D( _xi , x _s1 ), D( _xi , x _s2 ),..., D( _xi , x _si )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x; x _si represents the sample of the same category as i; the local metric matrix is A _S ; M _S =A _S ^T A _S ; i takes a positive integer ; D( _xi , x _si ) represents the distance between the input sample x _i and the samples of the same category as i in the target local cluster under the local metric matrix; i takes a positive integer; A _s ( _xi , x _si ) represents The distance between x _i and x _si after _AS matrix transformation.

In a second aspect, this embodiment provides a cardiovascular and cerebrovascular disease risk prediction device, including:

A collection acquisition module, used to obtain a sample set;

The sample set is determined according to multiple samples of the patient medical database set with labels; one sample includes: patient number, characteristics and characteristic data; the labels include: a first label and a second label; the first label Identify samples from patients with cardiovascular and cerebrovascular diseases; the second label identifies samples from patients with non-cardiovascular and cerebrovascular diseases;

A sample acquisition module, configured to acquire an input sample;

The input sample is formed by merging the medical health examination data and medical visit data of the patient to be predicted;

A matrix calculation module, configured to use the Cosine-Large Interval Nearest Neighbor COS-LMNN algorithm to perform metric learning to obtain the global metric matrix of the sample set;

The first local cluster determination module is configured to use a preset clustering algorithm to divide the samples in the sample set into a preset number of local clusters;

The first distance determination module is used to calculate the distance between the input sample and the samples in the sample set by using the cosine similarity algorithm according to the global metric matrix to form a first distance set;

The first sample determination module is configured to use the k-nearest neighbor algorithm to calculate the first K-value first neighboring samples of the input sample according to the preset first K value and the first distance set;

A second local cluster determination module, configured to determine the local cluster where the first adjacent sample is located;

A target local cluster determination module, configured to, among the local clusters where the first adjacent samples are located, select a local cluster whose number of first adjacent samples exceeds a first preset threshold as a target local cluster;

a local cluster division module, configured to divide the input samples into the target local clusters;

The second distance determination module is used to calculate the local metric matrix of the target local cluster learned according to the COS-LMNN algorithm, using the cosine similarity algorithm, and the distance between the input sample and the samples in the target local cluster to form the first Two-distance set;

The second sample determination module is configured to determine the second neighbors of the second K value of the input sample by using the k-nearest neighbor algorithm according to the preset second K value and the second distance set in the target local cluster sample;

A statistical module, configured to count the number of first labels and the number of second labels of the second adjacent samples;

A label determination module, configured to use the first label as the label of the input sample if the ratio of the number of the first label to the number of the second label exceeds the preset label threshold, otherwise the second label is used as the label of the input sample;

The patient sample determination module is used to determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases according to the label of the input sample;

A cardiovascular and cerebrovascular disease patient determination module is used to determine that the patient to be predicted in the input sample is a cardiovascular and cerebrovascular disease patient if the input sample is a sample of a patient with a cardiovascular and cerebrovascular disease;

The non-cardio-cerebrovascular disease patient determining module is used to determine that the patient to be predicted in the input sample is not a cardio-cerebrovascular disease patient if the input sample is not a sample of a cardio-cerebrovascular disease patient.

Optionally, a cardiovascular and cerebrovascular disease risk prediction device provided in this embodiment also includes:

A high-risk determination module, configured to determine whether the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient according to the patient's health follow-up data;

Hospitalization suggestion module, for if the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient, then make a suggestion for hospitalization of the patient to be predicted;

Adding a physical examination suggestion module, for if the patient to be predicted is not a high-risk cardiovascular and cerebrovascular disease patient, then make a suggestion to increase the frequency of physical examination for the patient to be predicted;

A health determination module, configured to determine whether the patient to be predicted is a healthy user according to the patient's health follow-up data;

A normal physical examination suggestion module, used to make suggestions for maintaining a normal physical examination frequency for the normal patient if the patient to be predicted is a healthy user;

A missed patient determination module, configured to mark the patient to be predicted as a missed patient if the patient to be predicted is not a healthy user, and add the feature data of the missed patient to the patient medical database set;

Among them, the missed diagnosis patients were patients with cardiovascular and cerebrovascular diseases.

Optionally, the collection acquisition module includes:

The sample deletion sub-module is used to delete samples whose sample missing value is greater than the first threshold according to multiple samples of the patient medical database set with labels;

The sample missing value is: the ratio of the number of missing features in a sample to the total number of features in the sample;

The feature deletion sub-module is used to search for multiple samples after deletion processing, and perform feature deletion processing for features whose feature missing value is greater than the second threshold;

The feature missing value is: in the same feature of multiple samples, the ratio of the number of features lacking feature data to the total number of the same feature;

The first feature sub-module is used to find the feature of missing feature data in multiple samples after feature deletion processing, as the first feature;

The missing value filling sub-module is used to fill the missing value of the feature data whose first feature is missing by using a multiple filling method;

The data classification sub-module is used to classify the characteristic data of the plurality of samples after the missing values are filled according to the data type, and obtain the classification result;

Wherein, the classification result includes: discrete feature data and continuous feature data;

The data processing sub-module is used to process the discrete feature data and continuous feature data corresponding to the data type according to the classification result;

The set update submodule is used to add the processed feature data corresponding to the discrete feature data and continuous feature data to the patient medical database set as the first database set;

Wherein, the discrete feature data and the continuous feature data are processed corresponding to the data type, including: performing one-hot encoding on the discrete feature data; standardizing the continuous feature data using the normalized z-score method;

The equalization processing sub-module is used to use the undersampling and SMOTE algorithm to perform unbalanced processing on the samples of the first database set to obtain the second database set;

The variance deletion submodule is used to calculate the variance of the same feature data in the second database set by using the variance analysis method, and delete the feature data whose variance value of the feature data is less than the preset variance threshold;

The weight calculation sub-module is used to use the relief algorithm to calculate the weight of each feature data after the feature data whose variance value is less than the preset variance threshold is deleted;

The set determination sub-module is used to determine the sample set by using the forward selection method according to the weight of each characteristic data and the second database set.

In yet another aspect of the implementation of the present invention, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete mutual communication through the communication bus;

memory for storing computer programs;

The processor is used to implement any one of the methods for predicting the risk of cardiovascular and cerebrovascular diseases described above when executing the program stored in the memory.

In yet another aspect of the implementation of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium stores instructions, and when it is run on a computer, it causes the computer to perform any one of the above-mentioned ones. A risk prediction method for cardiovascular and cerebrovascular diseases.

In yet another aspect of the implementation of the present invention, the embodiment of the present invention also provides a computer program product containing instructions, which, when run on a computer, enables the computer to perform any one of the above-mentioned cardiovascular and cerebrovascular disease risk predictions. method.

In the method and device for predicting the risk of cardiovascular and cerebrovascular diseases provided by the embodiments of the present invention, by obtaining a sample set of samples; dividing the samples in the sample set into a preset number of local clusters, according to the preset first K value and the set For the first distance set, use the k-nearest neighbor algorithm to calculate the first K-value first neighboring samples of the input sample, thereby determining the target local cluster, and after dividing the input sample into the target local cluster, calculate the input The distance between the sample and the sample in the target local cluster, thereby determining the second K-value second adjacent samples of the input sample; counting the number of first labels and the number of second labels of the second adjacent samples, thereby determining the label of the input sample , determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases; finally determine whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases. In this embodiment, considering the high similarity of the characteristic data of patients with cardiovascular and cerebrovascular diseases, the similarity distance between samples is calculated to determine the nearest sample of the input sample, so as to determine whether the patient to be predicted in the input sample is a cardiovascular and cerebrovascular disease patients, avoiding the impact of different sample feature data on the training prediction model. Therefore, the accuracy rate of predicting whether the patient to be predicted is a cardiovascular and cerebrovascular disease patient can be improved. Of course, implementing any product or method of the present invention does not necessarily need to achieve all the above-mentioned advantages at the same time.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a flow chart of a method for predicting risk of cardiovascular and cerebrovascular diseases according to an embodiment of the present invention;

Fig. 2 is the specific flowchart of step S101 in Fig. 1 in the embodiment of the present invention;

3 is a structural diagram of a cardiovascular and cerebrovascular disease risk prediction device according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In order to solve the problem of low accuracy in predicting whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases by using decision tree, logistic regression and artificial neural network algorithm to train the prediction model in the present embodiment, it can be understood that the heart and brain The characteristic data of patients with vascular diseases are similar. Therefore, the similarity distance between samples of patients with cardiovascular and cerebrovascular diseases can be used to obtain samples similar to the samples of patients to be predicted, so as to know whether the samples of patients to be predicted are cardiovascular and cerebrovascular. Disease patient samples, to determine whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases.

As shown in Figure 1, a method for predicting the risk of cardiovascular and cerebrovascular diseases provided by the embodiment of the present invention includes the following steps:

S101, acquire a sample set; the sample set is determined according to multiple samples of the patient medical database set with labels; one sample includes: patient number, characteristics and characteristic data; labels include: first label and second label; One label identifies samples from patients with cardiovascular and cerebrovascular diseases; the second label identifies samples from patients with non-cardio-cerebrovascular diseases;

For example, a sample includes: patient number 0001; features include: age, gender, city, occupation, family genetic history, disease history, diet, smoking habits, drinking habits, blood pressure, pulse, blood lipids, blood sugar, etc.; features Data include: Age: 50; Gender: Male; City: Wuhan; Occupation: Teacher; Family genetic history: None; Disease history: Hypertension; Dietary law: no breakfast, pasta or rice for lunch, barbecue for dinner; Smoking habit: one day Two cigarettes; drinking habits: at least once every three days; blood pressure: 100-145mmHg; pulse: 60-100 beats/min; serum total cholesterol: 2.9-5.17mmoi/l; serum triglycerides: 0.56-1.7mmoi/l; High-density lipoprotein cholesterol: 0,94-2.0mmoi/l; low-density lipoprotein cholesterol: 2.07-3.12i/l; blood sugar: fasting 7.8--9.0mmoL/L, etc.

It can be understood that the patient medical data database in this embodiment has multiple samples with labels set in a centralized manner, which is to combine the patient's medical examination data and medical visit data in advance to form a sample, or it can also be obtained every time a sample set , to combine the patient's medical examination data and medical visit data to form a sample. In view of the fact that the former can save time, this embodiment combines the patient's medical examination data and medical visit data in advance to form a sample, and then sets the label according to the patient's health return visit data. The patient's health follow-up data includes: the patient's label, the disease confirmed by the patient, and the data characteristics of the disease confirmed by the patient. For example: health follow-up data include: "stroke", "high blood pressure", "coronary heart disease", "high blood fat", "high blood sugar", "hemoptysis", "fainting", "cerebral infarction" and "heart failure "And many other cardiovascular and cerebrovascular diseases. The disease confirmed by the patient can be the disease diagnosed by the doctor or reflected by the patient's body in the prior art, or the disease confirmed by the patient's family members, and will not be repeated here.

The patient's medical examination data and medical visit data are the same as in the prior art. The medical examination data includes the patient's label and characteristic data, and the medical visit data includes the patient's label and basic information of the patient. Combine the patient's medical examination data and medical visit data according to the patient's label to form a sample, and then determine whether the sample is a sample of a patient with cardiovascular and cerebrovascular diseases according to the patient's label contained in the patient's health follow-up data and the disease confirmed by the patient, and give the sample Set tabs.

S102, obtaining an input sample; the input sample is formed by merging the medical health examination data and medical visit data of the patient to be predicted;

S103, using the cosine-large margin nearest neighbor COS-LMNN algorithm to perform metric learning to obtain a global metric matrix of the sample set;

The COS-LMNN algorithm in this embodiment is an algorithm that combines the cosine COS algorithm and the large-margin nearest neighbor LMNN algorithm to calculate the global metric matrix of the sample set.

S104, using a preset clustering algorithm to divide the samples in the sample set into a preset number of local clusters;

Wherein, the preset number is a number set according to industry experience, and the preset clustering algorithm may be the prior art k-means clustering algorithm, hierarchical clustering algorithm, SOM clustering algorithm, FCM clustering algorithm and the like. The preset number can be adjusted according to different clustering algorithms.

It can be understood that, in this embodiment, the samples in the sample set are classified and divided into different local clusters. For example, there are 7 samples in the sample set, namely A, B, C, D, E, F and G. The results are then local cluster 1: A and B; local cluster 2: C, F, G; local cluster 3: D and E.

S105. Calculate the distance between the input sample and the samples in the sample set by using the cosine similarity algorithm according to the global metric matrix to form a first distance set;

S106. According to the preset first K value and the first distance set, use the k-nearest neighbor algorithm to calculate and obtain the first K-value first neighboring samples of the input sample;

In this embodiment, the first K value is a preset value based on actual experience, and the value of the first K value is the same as the number of first adjacent samples calculated by using the k-nearest neighbor algorithm.

For example, the sample set includes: samples 1, 2, 3, and 4. Assuming that samples 1, 2, and 3 are samples of patients with cardiovascular and cerebrovascular diseases, the sample distance between samples 1, 2, and 3 can be 0. If the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases, and the second K value is set to 2, then the first K-value first adjacent samples of the input sample are samples 1 and 2, or 2 and 3, or 1 and 3.

S107, determine the local cluster where the first adjacent sample is located;

S108. In the local clusters where the adjacent samples are located, select a local cluster whose number of first adjacent samples exceeds a first preset threshold as a target local cluster;

S109, dividing the input samples into target local clusters;

S110, the local metric matrix of the target local cluster learned according to the COS-LMNN algorithm is calculated using a cosine similarity algorithm, and the distance between the input sample and the samples in the target local cluster is formed into a second distance set;

S111. In the target local cluster, according to the preset second K value and the second distance set, use the k-nearest neighbor algorithm to determine the second K-value second neighboring samples of the input sample;

In this embodiment, in order to determine the second adjacent samples of the input sample in the target local cluster, the number of the second adjacent samples is the same as the value of the second K value, and the determined second adjacent samples of the input sample can include multiple or one .

S112, counting the number of first labels and the number of second labels of the second adjacent samples;

S113, if the ratio of the number of the first label to the number of the second label exceeds the preset label threshold, use the first label as the label of the input sample, otherwise use the second label as the label of the input sample;

S114, according to the label of the input sample, determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases;

S115, if the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases, determine that the patient to be predicted in the input sample is a patient with cardiovascular and cerebrovascular diseases;

S116. If the input sample is not a sample of a patient with cardiovascular and cerebrovascular diseases, determine that the patient to be predicted in the input sample is not a patient with cardiovascular and cerebrovascular diseases.

Compared with the existing technology, data mining technology is used to mine the characteristics of case data of cardiovascular and cerebrovascular diseases, and the physical examination feature data and return visit data of all patients are combined into a training set, and decision trees, logistic regression and artificial neural networks are used to Algorithms to train predictive models. Then, the physical examination data of the patient to be predicted is used as an input sample, which is input into the trained prediction model, and whether the patient to be predicted is a cardiovascular and cerebrovascular disease patient is output.

Since the decision tree needs to analyze the physical examination data of all patients to be predicted in the process of training the prediction model, the sample data with the largest information gain of the physical examination data is used as the first node, and the other physical examination data are used as branches in order according to the information gain of the physical examination data. When the data Stop training when there is only one type of sample, and obtain a prediction model. When the physical examination data with the largest information gain is the physical examination data of patients with non-cardiovascular and cerebrovascular diseases, the prediction model trained by this method is affected by the sample data with the largest information gain in the physical examination data, and the prediction model trained by the decision tree predicts the results The accuracy rate is not high.

In the process of using the logistic regression algorithm to train the prediction model, it is necessary to find the minimum loss function and determine the prediction model. Since the minimum process of solving the loss function is easily affected by different sample data, the output of the prediction model is the probability that the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases, so it is not accurate to determine whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases.

In this embodiment, the acquired sample set of samples is obtained; the samples in the sample set are divided into a preset number of local clusters, and the first K-value first adjacent samples of the input samples are calculated to determine the target local cluster. By calculating the distance between the input sample and the sample in the target local cluster, thereby determining the second K-value second adjacent samples of the input sample; counting the number of first labels and the number of second labels of the second adjacent samples, so that Determine the label of the input sample, determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases; finally determine whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases. This embodiment does not need to train the prediction model. Considering the high similarity of the characteristic data of patients with cardiovascular and cerebrovascular diseases, the sample similarity distance is used to determine whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases, and the decision tree is avoided in training the prediction model. In the process, the sample data with the largest information gain of the physical examination data is affected. In this embodiment, it is not necessary to use the logistic regression algorithm to solve the loss function to train the prediction model, so as to avoid the influence of different sample data on the minimum process of solving the loss function, resulting in inaccurate prediction models trained. Therefore, the accuracy rate of predicting whether the patient to be predicted is a cardiovascular and cerebrovascular disease patient can be improved.

Optionally, in an embodiment of the risk prediction method for cardiovascular and cerebrovascular diseases of the present invention, in S115, if the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases, it is determined that the patient to be predicted in the input sample is a patient with cardiovascular and cerebrovascular diseases After the step, the method also includes:

Step 1: Determine whether the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient according to the patient's health follow-up data;

The feature data in the health follow-up data exceeds the normal index, the sign information is abnormal, and the symptoms are abnormal. Therefore, it can be judged that the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient. For example, blood pressure exceeds the standard, hemoglobin exceeds the standard, and symptoms are abnormal, such as fainting, coughing and hemoptysis.

Step 2: If the patient to be predicted is a patient with high-risk cardiovascular and cerebrovascular diseases, then recommend hospitalization for the patient to be predicted;

According to the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient, according to the characteristic data of the patient, the hospitalization days and treatment plan corresponding to the characteristic data of the patient are provided to the patient.

In this embodiment, a treatment database for cardiovascular and cerebrovascular patients is preliminarily established. The treatment database for cardiovascular and cerebrovascular patients includes: characteristic data of patients, days of hospitalization corresponding to the characteristic data, and treatment plans. The treatment plan includes: the amount and frequency of insulin injection, the amount and frequency of taking antihypertensive drugs, physical exercise or whether surgical treatment is needed, etc.

For example: blood pressure of patients with high-risk cardiovascular and cerebrovascular diseases: 120-155mmHg; blood sugar: fasting 7.8--9.0mmoL/L corresponds to 20 days of hospitalization, and the treatment plan corresponding to the characteristic data is to inject 1U of insulin per day.

This embodiment saves time for doctors to diagnose and advise, and saves medical resources. It is judged that the patient is a high-risk cardiovascular and cerebrovascular disease patient, and a suggestion for hospitalization of the patient to be predicted is given.

Step 3: If the patient to be predicted is not a high-risk cardiovascular and cerebrovascular disease patient, make suggestions to increase the frequency of physical examination for the patient to be predicted.

It can be understood that in this embodiment, if the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases, if the patient's health follow-up data records the characteristic data of the patient to be predicted to belong to a certain number of characteristic data of patients with cardiovascular and cerebrovascular diseases, the to-be-predicted It is predicted that the patient's signs and symptoms are not abnormal, and the patient is not a high-risk cardiovascular and cerebrovascular disease patient. According to the range of characteristic data, the frequency of physical examination corresponding to the range of characteristic data may be suggested to the patient.

For example: the blood pressure range of 100 patients with cardiovascular and cerebrovascular diseases is: 100-145mmHg, and the blood pressure of the patient to be predicted is: 102-140mmHg. Patients with cerebrovascular disease. If the patient's previous physical examination frequency is once a month, the blood pressure value range in the patient's characteristic data is 102-140mmHg. Assuming that the physical examination frequency corresponding to the patient's characteristic data is twice a month, it is recommended that the patient undergo a physical examination twice a month. This embodiment saves time for doctors to diagnose and advise, and saves medical resources.

Optionally, in an embodiment of the method for predicting the risk of cardiovascular and cerebrovascular diseases of the present invention, in S115, after the step of determining that the patient to be predicted in the input sample is not a patient with cardiovascular and cerebrovascular diseases, it also includes:

Step 1: Determine whether the patient to be predicted is a healthy user according to the patient's health follow-up data;

Among them, the healthy user is a patient whose characteristic data are all within the standard range specified by the medical characteristic data. For example: the normal blood pressure stipulated by medicine: 80-90/120-140mmHg, if the health follow-up data shows that the blood pressure of the patient to be predicted is 82/125mmHg, and other characteristic data of the patient are within the standard range stipulated by various medical characteristic data, the A patient is a healthy user.

Step 2: If the patient to be predicted is a healthy user, make suggestions for normal patients to maintain normal physical examination frequency;

It can be understood that, if the patient to be predicted is a healthy user, the frequency of physical examination of the healthy user corresponds to the data characteristics of the patient. It is recommended that the patient maintain the same number of physical examinations as the previous physical examination frequency. For example: the patient's previous physical examination frequency was once a month, and it is recommended to maintain the monthly physical examination frequency. In this embodiment, healthy users are selected and appropriate suggestions are given, which saves time for doctors to diagnose and recommend, and reduces medical expenses for patients.

Step 3: If the patient to be predicted is not a healthy user, mark the patient to be predicted as a missed diagnosis patient, and add the characteristic data of the missed diagnosis patient to the patient medical database set;

Among them, the missed diagnosis patients were patients with cardiovascular and cerebrovascular diseases.

It can be understood that if the patient to be predicted is not a healthy user, it can be determined whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases or not according to the patient's health follow-up data. If the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases, mark the patient as a missed diagnosis patient, and add the characteristic data of the patient to the patient medical database set to prevent the same type of patients to be predicted from predicting whether they are patients with cardiovascular and cerebrovascular diseases , misprediction occurs, and the accuracy of predicting whether the patient to be predicted is a patient with cardiovascular and cerebrovascular diseases is improved.

Optionally, in an embodiment of the risk prediction method for cardiovascular and cerebrovascular diseases of the present invention, the first label identifies samples of patients with cardiovascular and cerebrovascular diseases, including:

Step 1: Determine the identification information of patients with cardiovascular and cerebrovascular diseases according to the collected patient health follow-up data;

The patient's health follow-up data includes: the patient's number, characteristics, characteristic data and confirmed symptoms; identification information includes: confirmed symptoms, confirmed symptoms corresponding characteristics and characteristic data;

Step 2: According to the identification information of patients with cardiovascular and cerebrovascular diseases, determine the samples of patients with cardiovascular and cerebrovascular diseases in the medical database set;

Step 3: Set the first label on the samples of patients with cardiovascular and cerebrovascular diseases.

In this embodiment, through the patient's health follow-up data, it is distinguished that the medical database centrally determines samples of patients with cardiovascular and cerebrovascular diseases, and the first label is set to save time for determining the label of the input sample.

Optionally, in an embodiment of the method for predicting the risk of cardiovascular and cerebrovascular diseases of the present invention, the second label identifies samples of patients with non-cardio-cerebrovascular diseases, including:

A second label is set for samples other than samples of patients with cardiovascular and cerebrovascular diseases.

In this embodiment, the health follow-up data of one month after the user's physical examination can be used. Labels are set for samples of cardiovascular and cerebrovascular diseases and samples of non-cardiovascular and cerebrovascular diseases in the health follow-up data. Labels include: letters, numbers, symbols, and more. For example: health follow-up data include: "stroke", "hypertension", "coronary heart disease", "hyperlipidemia", "hyperglycemia", "cerebral infarction" and "heart failure" and other description settings for cardiovascular and cerebrovascular diseases The label is a positive sample, and a new category "label" field is added, and the label "1" is set. All samples that are not positive samples are taken as negative samples, and the label "0" is set to join the medical database set.

In this embodiment, through the patient's health follow-up data, the medical database is used to distinguish the non-cardiovascular and cerebrovascular disease patient samples, and set the second label to save time for determining the label of the input sample.

Optionally, in an embodiment of a cardiovascular and cerebrovascular disease risk prediction method of the present invention, S101 acquires a sample set, including:

S201, according to the multiple samples of the patient medical database set with labels, delete the samples whose sample missing value is greater than the first threshold;

The sample missing value is: the ratio of the number of missing features in a sample to the total number of features in the sample;

Among them, the first threshold is a value artificially specified based on industry experience. The following example illustrates the missing value of the sample. For example, if a sample contains 10 features and 7 features are missing feature data, then the ratio of the number of missing features in the sample to the total number of features in the sample is Assume that the specified first threshold is The sample is then treated as a sample deletion.

In this embodiment, the purpose of deleting samples whose missing values are greater than the first threshold is to reduce samples with less characteristic data in the patient medical database, improve the quality of samples in the patient medical database, and save time for subsequent processing.

S202, searching among the multiple samples after deletion processing, performing feature deletion processing for features whose feature missing value is greater than the second threshold;

In this embodiment, features whose feature missing value is greater than the second threshold are processed for feature deletion; the purpose is to reduce the less feature data in the samples of the patient medical database set, improve the quality of feature data in the samples of the patient medical database set, and Save time for subsequent processing.

The feature missing value is: the ratio of the number of features lacking feature data to the total number of the same feature in the same feature of multiple samples;

Wherein, the second threshold is a numerical value artificially specified based on industry experience. The following example illustrates the missing value of a feature, for example, there is the same feature in 10 samples: pulse. The number of pulse features lacking feature data is 7, and the total number of pulse features is 10. Assume that the specified first threshold is The pulse feature is then processed as feature deletion.

In this embodiment, the features whose feature missing value is greater than the second threshold are used for feature deletion processing. The purpose is to reduce the less feature data in the samples of the patient medical database set, improve the quality of the feature data in the samples of the patient medical database set, and for Subsequent processing saves time.

S203, searching for the feature of the missing feature data in the multiple samples after feature deletion processing, as the first feature;

S204, using a multiple filling method to fill in missing values for the feature data whose first feature is missing;

Among them, the module constructed using the multiple imputation method in IBM SPSS statistics 23 is used to fill in the missing values. For example, there are two sample blood pressure characteristic data missing, and the module constructed using the multiple imputation method is based on the characteristic data of the patient: age: 50; Blood lipids: 1. Serum total cholesterol 2.9-5.17mmoi/l; 2. Serum triglycerides 0.56-1.7mmoi/l; 3. High-density lipoprotein cholesterol 0.94-2.0mmoi/l; 4. Low-density lipoprotein cholesterol 2.07- 3.12i/l; blood sugar: 7.8--9.0mmoL/L on an empty stomach, fill in the blood pressure characteristic data in the two samples with a value of 100-145mmHg, the specific filling method is the same as that of the prior art, and will not be repeated here.

In this embodiment, the missing feature data of multiple samples is filled with missing values, which can improve the quality of the samples, so as to improve the quality of the obtained sample set.

S205, according to the data type, classify the characteristic data of the multiple samples after the missing value is filled, and obtain the classification result;

Among them, the classification results include: discrete feature data and continuous feature data;

S206, according to the classification result, process the discrete feature data and the continuous feature data corresponding to the data type;

S207, adding the processed feature data corresponding to the discrete feature data and the continuous feature data into the patient medical database set as the first database set;

Among them, the discrete feature data and continuous feature data are processed corresponding to the data type, including: performing one-hot encoding on the discrete feature data; and standardizing the continuous feature data using the normalized z-score method;

The data types of the characteristic data of the patient include: discrete characteristic data and continuous characteristic data. For example, blood pressure data and heartbeat data are continuous types, and age data are discrete types.

For example, for discrete feature data, write code that applies one-hot encoding of the feature data. Taking the encoding of "age" as an example, first, divide the age feature into equal frequency segments according to the number of samples, and divide it into "76 and above", "66-75", "55-65", "46-55", " There are seven intervals: 36-45", "26-35", and "below 25". If a person's age is 30 years old, the age value after one-hot encoding is 0000010. Other discrete features are similar to age features, such as gender, city, occupation, family genetic history, disease history, eating habits, smoking habits, drinking habits, weekly exercise patterns, etc., all undergo one-hot encoding conversion.

It can be understood that, in this embodiment, the normalized z-score method is used to standardize the continuous feature data in the same way as in the prior art, and will not be repeated here. In this embodiment, according to different data types, discrete feature data and continuous feature data are processed correspondingly, so as to avoid consistent data results caused by using the same method to process data, which can improve the accuracy of data processing.

S208, using the undersampling and SMOTE algorithm to perform unbalanced processing on the samples of the first database set to obtain the second database set;

In this embodiment, the unbalanced processing is performed on the samples of the first database set to make the distribution of samples of the same type relatively uniform, so as to obtain an accurate second database set.

S209, using the variance analysis method to calculate the variance of the same characteristic data in the second database set, and delete the characteristic data whose variance value of the characteristic data is smaller than the preset variance threshold, and obtain the third database set;

This embodiment chooses to delete the feature data whose variance value of the feature data is less than the preset variance threshold, which can reduce the data with small difference in the sample feature data, and delete the feature data whose variance value of the feature data is less than the preset variance threshold. The second database Set, as the third database set. It can be understood that the greater the difference value, the greater the difference of the samples, and the higher the accuracy of distinguishing the samples of cardiovascular and cerebrovascular diseases from non-cardio and cerebrovascular diseases.

S210, use the relief algorithm to calculate, delete the weight of each feature data after the feature data whose variance value is smaller than the preset variance threshold;

S211. According to the weight of the feature data and the score value corresponding to the weight of the feature data, delete the feature data and the corresponding feature whose score value is less than the preset score threshold, and obtain the fourth database set;

In this embodiment, a weight database is established in advance, and the weight database includes: the weight of the feature data and the score value corresponding to the weight of the feature data. According to the weight of each feature data, the score value corresponding to the weight of each feature data is searched in the database, and each feature data is scored. Delete the feature data and corresponding features whose score values in the third database set do not exceed the score threshold, and the score threshold is a value set according to industry experience.

S212. Determine a sample set by using a forward selection method according to the fourth database set.

It can be understood that, in the process of using the forward selection method to determine the sample set, the evaluation function can be used to evaluate the feature data corresponding to each feature in the fourth database set, and determine the evaluation function value of the feature data corresponding to each feature.

In some examples, for the feature data corresponding to the same feature, the samples corresponding to the feature data corresponding to the feature with the same evaluation function value can be used as a model sample set. The model sample set includes multiple samples, and the multiple samples include: at least one The same feature and feature data corresponding to the feature; then evaluate each model sample set, and finally select the model sample set with the highest evaluation function value as the sample set. Evaluating each model sample set may include: calculating the average value of all feature data evaluation function values in the model sample set, or choosing to calculate the average value of the feature data related to cerebrovascular diseases in the model sample set, and evaluating each model sample set can also use the existing The method of technology evaluation set will not be repeated here.

The following example illustrates: Suppose there are 3 characteristic data: blood pressure: 100-145mmHg; pulse: 60-100 beats/min; blood sugar: fasting 7.8--9.0mmoL/L. The evaluation function values of the three characteristic data are 64, 78, and 12 respectively; select the samples where the pulse rate is 60-100 beats/min to form the model sample set 1; select the samples where the blood pressure: 100-145mmHg is located to form the model sample set 2 ;Choose the blood sugar: fasting 7.8--9.0mmoL/L samples to form the model sample set 3; The average value is 45 points, the average value of the characteristic data evaluation function value in model sample set 3 is 65 points, and model sample set 3 is used as the sample set.

In this embodiment, the quality of the samples is improved by preprocessing the samples of the patient medical database set and the feature data in the samples, so the quality of the sample sets can be improved.

Optionally, in step S105, according to the global metric matrix, use the cosine similarity algorithm to calculate the distance between the input sample and the samples in the sample set to form a first distance set, including:

According to the global metric matrix, using the cosine similarity algorithm formula to calculate the distance between the input sample and the samples in the sample set to form a first distance set;

Wherein, the cosine similarity algorithm formula is:

The first distance set D1 includes: {D( _xi ,x ₁ ), D( _xi ,x ₂ ), D( _xi ,x ₃ ),...,D( _xi ,x _j )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x _i ; the sample set is X; the global metric matrix is A; M= ^AT A; j represents the sample number in the sample set; x _j represents the sample The jth sample in the set; i and j take positive integers; D(s _i , x _j ) represents the distance between the input sample x _i and the jth sample in the X set under the global metric matrix; A(x _i , x _j ) Represents the distance between x _i and x _j after A matrix transformation.

Optionally, in step S110, the local metric matrix of the target local cluster learned according to the COS-LMNN algorithm is calculated using the cosine similarity algorithm, and the distance between the input sample and the samples in the target local cluster forms a second distance set, including:

Use the cosine similarity algorithm formula to calculate the distance between the input sample and the sample in the sample set;

Among them, the cosine similarity algorithm formula is:

The second distance set D2 includes: {D(x _i , x _s1 ), D(x _i , x _s2 ), . . . , D(x _i , x _si )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x; x _si represents the sample of the same category as i; the local metric matrix is A _S ; M _S =A _S ^T A _S ; i takes a positive integer ; D( _xi , x _si ) represents the distance between the input sample x _i and the samples of the same category as i in the target local cluster under the local metric matrix; i takes a positive integer; A _S ( _xi , x _si ) represents The distance between x _i and x _si after _AS matrix transformation.

As shown in Figure 3, a cardiovascular and cerebrovascular disease risk prediction device provided by an embodiment of the present invention includes:

A set acquisition module 301, configured to acquire a sample set;

The sample set is determined according to multiple samples of the patient medical database set with labels; one sample includes: the patient's number, characteristics and characteristic data; the labels include: the first label and the second label; the first label identifies the heart Samples from patients with cerebrovascular diseases; the second label identifies samples from patients with non-cerebrovascular diseases;

A sample acquisition module 302, configured to acquire an input sample;

The input sample is composed of the medical health examination data and medical visit data of the patient to be predicted;

The matrix calculation module 303 is used to use the cosine-large interval nearest neighbor COS-LMNN algorithm to perform metric learning to obtain the global metric matrix of the sample set;

The first local cluster determination module 304 is configured to use a preset clustering algorithm to divide the samples in the sample set into a preset number of local clusters;

The first distance determination module 305 is used to calculate the distance between the input sample and the samples in the sample set by using the cosine similarity algorithm according to the global metric matrix to form a first distance set;

The first sample determination module 306 is configured to use the k-nearest neighbor algorithm to calculate the first K-value first neighboring samples of the input sample according to the preset first K value and the first distance set;

The second local cluster determination module 307 is configured to determine the local cluster where the first adjacent sample is located;

A target local cluster determination module 308, configured to select, among the local clusters where the adjacent samples are located, a local cluster whose number of first adjacent samples exceeds a first preset threshold, as the target local cluster;

A local cluster division module 309, configured to divide input samples into the target local clusters;

The second distance determination module 310 is used to learn the local metric matrix of the target local cluster according to the COS-LMNN algorithm, and use the cosine similarity algorithm to calculate the distance between the input sample and the samples in the target local cluster to form a second distance set;

The second sample determination module 311 is used to determine the second K-value second adjacent samples of the input sample by using the k-nearest neighbor algorithm according to the preset second K value and the second distance set in the target local cluster;

A statistical module 312, configured to count the number of first labels and the number of second labels of the second adjacent samples;

A label determination module 313, configured to use the first label as the label of the input sample if the ratio of the number of the first label to the number of the second label exceeds the preset label threshold, otherwise the second label is used as the label of the input sample;

The patient sample determination module 314 is used to determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases according to the label of the input sample;

Cardiovascular and cerebrovascular disease patient determination module 315 is used to determine that the patient to be predicted in the input sample is a cardiovascular and cerebrovascular disease patient if the input sample is a sample of a patient with cardiovascular and cerebrovascular disease;

The non-cardiovascular and cerebrovascular disease patient determination module 316 is configured to determine that the patient to be predicted in the input sample is not a cardiovascular and cerebrovascular disease patient if the input sample is not a sample of a cardiovascular and cerebrovascular disease patient.

Optionally, a cardiovascular and cerebrovascular disease risk prediction device provided in the embodiments of the present invention also includes:

The high-risk determination module is used to determine whether the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient according to the patient's health follow-up data;

The hospitalization suggestion module is used for if the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient, then the patient to be predicted is recommended to be hospitalized;

Add a physical examination suggestion module, which is used to make suggestions to increase the frequency of physical examination for the predicted patient if the predicted patient is not a high-risk cardiovascular and cerebrovascular disease patient;

A health determination module is used to determine whether the patient to be predicted is a healthy user according to the patient's health return visit data;

A normal physical examination suggestion module, used for making suggestions on maintaining a normal physical examination frequency for the normal patient if the patient to be predicted is a healthy user;

A missed diagnosis module is used for if the patient to be predicted is not a healthy user, then the patient to be predicted is marked as a missed diagnosis patient, and the characteristic data of the missed diagnosis patient is added to the patient medical database set;

Among them, the missed diagnosis patients were patients with cardiovascular and cerebrovascular diseases.

Optionally, the collection acquisition module 301 includes:

The sample deletion sub-module is used to delete samples whose sample missing value is greater than the first threshold according to multiple samples of the patient medical database set with labels;

The sample missing value is: the ratio of the number of missing features in a sample to the total number of features in the sample;

The feature deletion sub-module is used to search in multiple samples after deletion processing, and perform feature deletion processing for features whose feature missing value is greater than the second threshold;

The feature missing value is: in the same feature of multiple samples, the ratio of the number of features lacking feature data to the total number of the same feature;

The first feature sub-module is used to find the feature of missing feature data in multiple samples after feature deletion processing, as the first feature;

The missing value filling submodule is used to fill the missing value of the feature data whose first feature is missing by using a multiple filling method;

The data classification sub-module is used to classify the characteristic data of the plurality of samples after the missing values are filled according to the data type, and obtain the classification result;

Wherein, the classification result includes: discrete feature data and continuous feature data;

The data processing sub-module is used to process the discrete feature data and continuous feature data corresponding to the data type according to the classification result;

The set update submodule is used to add the processed feature data corresponding to the discrete feature data and continuous feature data to the patient medical database set as the first database set;

Wherein, the discrete feature data and the continuous feature data are processed corresponding to the data type, including: performing one-hot encoding on the discrete feature data; standardizing the continuous feature data using the normalized z-score method;

The equalization processing sub-module is used to use the undersampling and SMOTE algorithm to perform unbalanced processing on the samples of the first database set to obtain the second database set;

The variance deletion sub-module is used to calculate the variance of the same characteristic data in the second database set using the variance analysis method, delete the characteristic data whose variance value of the characteristic data is less than the preset variance threshold, and obtain the third database set;

A weight calculation submodule, configured to use a relief algorithm to calculate the weight of each characteristic data in the third database set;

The score deletion sub-module is used to delete the feature data and corresponding features whose score value is less than the preset score threshold in the third database set according to the weight of the feature data and the score value corresponding to the weight of the feature data, and obtain the fourth database set;

The set determining submodule is used to determine the sample set according to the fourth database set by using a forward selection method.

A cardiovascular and cerebrovascular disease risk prediction device of the present embodiment also includes:

The high-risk determination module is used to determine whether the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient according to the patient's health follow-up data;

The hospitalization suggestion module is used for if the patient to be predicted is a high-risk cardiovascular and cerebrovascular disease patient, then the patient to be predicted is recommended to be hospitalized;

Add a physical examination suggestion module, which is used to make suggestions to increase the frequency of physical examination for the predicted patient if the predicted patient is not a high-risk cardiovascular and cerebrovascular disease patient;

A health determination module is used to determine whether the patient to be predicted is a healthy user according to the patient's health return visit data;

The normal physical examination suggestion module is used to make suggestions on maintaining normal physical examination frequency for normal patients if the patient to be predicted is a healthy user;

The missed patient determination module is used to mark the patient to be predicted as a missed patient if the patient to be predicted is not a healthy user, and add the feature data of the missed patient to the patient medical database set;

Among them, the missed diagnosis patients were patients with cardiovascular and cerebrovascular diseases.

Optionally, the first distance determination module is specifically configured to: calculate the distance between the input sample and the samples in the sample set by using the cosine similarity algorithm formula according to the global metric matrix to form the first distance set;

Among them, the cosine similarity algorithm formula is:

The first distance set D1 includes: {D( _xi ,x ₁ ), D( _xi ,x ₂ ), D( _xi ,x ₃ ),...,D( _xi ,x _j )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x _i ; the sample set is X; the global metric matrix is A; M= ^AT A; j represents the sample number in the sample set; x _j represents the sample The j-th sample in the set; i and j take positive integers; D( _xi , x _j ) represents the distance between the input sample x _i and the j-th sample in the X set under the global metric matrix; A( _xi , x _j ) Represents the distance between x _i and x _j after A matrix transformation.

Optionally, the second distance determination module is specifically used for:

According to the local metric matrix of the target local cluster learned by the COS-LMNN algorithm, using the cosine similarity algorithm formula to calculate the distance between the input sample and the samples in the sample set to form a second distance set;

Among them, the cosine similarity algorithm formula is:

The second distance set D2 includes: {D(x _i , x _s1 ), D(x _i , x _s2 ), . . . , D(x _i , x _si )};

Among them, i represents the label of the input sample, x _i represents the i-th input sample as x; x _si represents the sample of the same category as i; the local metric matrix is A _S ; M _S =A _S ^T A _S ; i takes a positive integer ; D( _xi , x _si ) represents the distance between the input sample x _i and the samples of the same category as i in the target local cluster under the local metric matrix; i takes a positive integer; A _s ( _xi , x _si ) represents The distance between x _i and x _si after _AS matrix transformation.

The embodiment of the present invention also provides an electronic device, as shown in FIG. complete the mutual communication,

Memory 403, used to store computer programs;

When the processor 401 is used to execute the program stored on the memory 403, the following steps are implemented:

Obtain a sample set; the sample set is determined based on multiple samples of the patient medical database set with labels; a sample includes: the patient's number, characteristics and characteristic data; the labels include: the first label and the second label; the first label Identify samples from patients with cardiovascular and cerebrovascular diseases; the second label identifies samples from patients with non-cardiovascular and cerebrovascular diseases;

Obtain an input sample; the input sample is composed of the medical health examination data and medical visit data of the patient to be predicted;

Use the cosine-large interval nearest neighbor COS-LMNN algorithm for metric learning to obtain the global metric matrix of the sample set;

Use a preset clustering algorithm to divide the samples in the sample set into a preset number of local clusters;

According to the global metric matrix, using the cosine similarity algorithm to calculate the distance between the input sample and the samples in the sample set to form a first distance set;

According to the preset first K value and the first distance set, use the k-nearest neighbor algorithm to calculate the first K-value first neighboring samples of the input sample;

determining the local cluster in which the first neighboring sample is located;

In the local clusters where the adjacent samples are located, select a local cluster whose number of first adjacent samples exceeds a first preset threshold as a target local cluster;

dividing input samples into said target local clusters;

The local metric matrix of the target local cluster learned according to the COS-LMNN algorithm is calculated using a cosine similarity algorithm, and the distance between the input sample and the samples in the target local cluster forms a second distance set;

In the target local cluster, according to the preset second K value and the second distance set, use the k-nearest neighbor algorithm to determine the second K-value second adjacent samples of the input sample;

counting the number of the first label and the number of the second label of the second adjacent sample;

If the ratio of the number of the first label to the number of the second label exceeds the preset label threshold, the first label is used as the label of the input sample, otherwise the second label is used as the label of the input sample;

Determine whether the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases according to the label of the input sample;

If the input sample is a sample of a patient with cardiovascular and cerebrovascular diseases, it is determined that the patient to be predicted in the input sample is a patient with cardiovascular and cerebrovascular diseases;

If the input sample is not a sample of a patient with cardiovascular and cerebrovascular diseases, it is determined that the patient to be predicted in the input sample is not a patient with cardiovascular and cerebrovascular diseases.

The communication bus mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus or the like. The communication bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

The communication interface is used for communication between the electronic device and other devices.

The memory may include a random access memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far away from the aforementioned processor.

Above-mentioned processor can be general-purpose processor, comprises central processing unit (Central Processing Unit, CPU), network processor (Network Processor, NP) etc.; Can also be Digital Signal Processor (Digital Signal Processing, DSP), ASIC (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

In yet another embodiment provided by the present invention, a computer-readable storage medium is also provided. Instructions are stored in the computer-readable storage medium. When the computer-readable storage medium is run on a computer, it causes the computer to execute any one of the above-mentioned embodiments. A method for predicting the risk of cardiovascular and cerebrovascular diseases.

In yet another embodiment provided by the present invention, a computer program product containing instructions is also provided, and when it is run on a computer, it makes the computer execute the cardiovascular and cerebrovascular disease risk control described in any one of the above embodiments. method of prediction.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. any such actual relationship or order exists between them. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

Each embodiment in this specification is described in a related manner, the same and similar parts of each embodiment can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, as for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for relevant parts, please refer to part of the description of the method embodiment.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention are included in the protection scope of the present invention.

Claims (10)

1. a kind of cardiovascular and cerebrovascular disease Risk Forecast Method, which is characterized in that the method includes：

Obtain sample set；Determined by multiple samples of the sample set according to the patient medical data library collection for setting up label； One sample includes：Number, feature and the characteristic of patient；The label includes：First label and the second label；First mark Label mark Patients with Cardiovascular/Cerebrovascular Diseases sample；The non-Patients with Cardiovascular/Cerebrovascular Diseases sample of second tag identifier；

Obtain an input sample；The input sample is by the medical treatment ＆ health physical examination data of patient to be predicted and the medical data of medical treatment Merge and constitutes；

Metric learning is carried out using cosine-large-spacing nearest-neighbors COS-LMNN algorithms, obtains the global measurement of the sample set Matrix；

Using preset clustering algorithm, the sample in sample set is divided into the local cluster of preset quantity；

According to the global metric matrix input sample and sample in the sample set are calculated using cosine similarity algorithm This distance forms the first distance set；

The first K of input sample is calculated using k nearest neighbor algorithms according to preset first K values and first distance set Value first is adjacent to sample；

Described first is determined adjacent to the local cluster where sample；

Described first in the local cluster where sample, selection first is more than the first predetermined threshold value adjacent to the quantity of sample Local cluster, as target part cluster；

The input sample is included in target part cluster；

According to the Local Metric matrix for the target part cluster that COS-LMNN algorithms learn, cosine similarity algorithm is used Calculate, the input sample in the cluster of the target part at a distance from sample, composition second distance set；

In the cluster of the target part, determined using k nearest neighbor algorithms according to preset 2nd K values and the second distance set 2nd K values second of the input sample are adjacent to sample；

First label number and second label number of the statistics second adjacent to sample；

If second is more than default label threshold value adjacent to the ratio of the first label number and the second label number of sample, by the Label of one label as input sample, otherwise using the second label as the label of input sample；

According to the label of the input sample, determine input sample whether be Patients with Cardiovascular/Cerebrovascular Diseases sample；

If input sample is the sample of Patients with Cardiovascular/Cerebrovascular Diseases, it is determined that the patient to be predicted in input sample is heart and brain blood Pipe Disease；

If input sample is not the sample of Patients with Cardiovascular/Cerebrovascular Diseases, it is determined that the patient to be predicted in input sample is not the heart Cerebrovascular patients.

2. according to the method described in claim 1, it is characterized in that,

After the step of patient to be predicted in the determining input sample is Patients with Cardiovascular/Cerebrovascular Diseases, the method is also wrapped It includes：

Determine whether the patient to be predicted is high-risk Patients with Cardiovascular/Cerebrovascular Diseases according to the health follow-up data of patient；

If the patient to be predicted is high-risk Patients with Cardiovascular/Cerebrovascular Diseases, the patient to be predicted is built as hospitalization View；

If the patient to be predicted is not high-risk Patients with Cardiovascular/Cerebrovascular Diseases, the patient to be predicted is made to increase physical examination frequency Secondary suggestion；

After the step of patient to be predicted in the determining input sample is not Patients with Cardiovascular/Cerebrovascular Diseases, the method is also Including：

Determine whether the patient to be predicted is healthy user according to the health follow-up data of patient；

If the patient to be predicted is healthy user, make the suggestion for keeping the regular inspection frequency to the normal patient；

It is to fail to pinpoint a disease in diagnosis patient by the patient indicia to be predicted, by the leakage if the patient to be predicted is not healthy user Patient medical data library collection is added in the characteristic for examining patient；

Wherein, it is Patients with Cardiovascular/Cerebrovascular Diseases to fail to pinpoint a disease in diagnosis patient.

3. according to the method described in claim 1, it is characterized in that,

The first tag identifier Patients with Cardiovascular/Cerebrovascular Diseases sample, including：

According to the health follow-up data of the patient collected, the identification information of Patients with Cardiovascular/Cerebrovascular Diseases is determined；

The health follow-up data of the patient include：Number, feature, characteristic and the confirmation illness of patient；The mark letter Breath includes：Confirm illness, confirm the corresponding feature of illness and characteristic；

According to the identification information of Patients with Cardiovascular/Cerebrovascular Diseases, is concentrated in the medical data base and determine Patients with Cardiovascular/Cerebrovascular Diseases sample This；

By the sample of the Patients with Cardiovascular/Cerebrovascular Diseases, the first label is set；

The non-Patients with Cardiovascular/Cerebrovascular Diseases sample of second tag identifier, including：

By other samples in addition to the Patients with Cardiovascular/Cerebrovascular Diseases sample, the second label is set.

4. according to the method described in claim 1, it is characterized in that, the acquisition sample set, including：

According to multiple samples of the patient medical data library collection of setting label, the sample that sample missing values are more than to first threshold is made Sample delete processing；

The sample missing values are：The ratio of the feature quantity lacked in one sample and feature total quantity in the sample；

It is searched in a plurality of sample after delete processing, the feature that feature missing values are more than second threshold makees feature delete processing；

The feature missing values are：In the same feature of a plurality of sample, the feature quantity for the data that lack in individuality and same feature are total The ratio of quantity；

A plurality of sample after making feature delete processing searches the feature of missing characteristic, as fisrt feature；

Using Multiple Imputation, the characteristic of fisrt feature missing is filled up as missing values；

According to data type, the characteristic of a plurality of sample after being filled up to missing values, which is done, classifies, and obtains classification results；

Wherein, the classification results include：Discrete features data and continuous characteristic；

The discrete features data and continuous characteristic are made into processing corresponding with data type according to classification results；

The discrete features data and continuous characteristic are done into corresponding treated characteristic, the patient doctor is added Data base set is treated, as first database collection；

Wherein, by the discrete features data and continuous characteristic, make processing corresponding with data type, including：To discrete Characteristic carries out one-hot coding；To continuous characteristic, it is standardized using just too standardization z-score methods；

Using lack sampling and SMOTE algorithms, unbalanced processing is carried out to the sample of the first database collection, obtains the second data Library collection；

The variance of the same characteristic of second centralized database is calculated using method of analysis of variance, deletes characteristic variance Value is less than the characteristic of default variance threshold values, obtains third data base set；

The weight of each characteristic of third centralized database is calculated using relief algorithms；

According to the weight of characteristic fractional value corresponding with the weight of characteristic, third centralized database fractional value is less than The characteristic of preset fraction threshold value and corresponding feature are deleted, and the 4th data base set is obtained；

According to the 4th data base set sample set is determined using forward selection procedures.

5. according to the method described in claim 1, it is characterized in that,

It is described according to the global metric matrix, using cosine similarity algorithm calculate the input sample in the sample set The distance of sample forms the first distance set, including：

According to the global metric matrix, calculated in the input sample and the sample set using cosine similarity algorithmic formula The distance of sample forms the first distance set；

Wherein, the cosine similarity algorithmic formula is：

The first distance set D1 includes：{D(x_i,x₁), D (x_i,x₂), D (x_i,x₃) ..., D (x_i,x_j)}；

Wherein, i represents the label of input sample, x_iI-th of input sample is represented as x_i；Sample set is X；Global metric matrix is A；M=A^TA；The sample number that j representative samples are concentrated；x_jRepresentative sample concentrates j-th of sample；I and j takes positive integer；D(x_i, x_j) represent the input sample x under global metric matrix_iIt is concentrated at a distance from j-th of sample with X；A(x_i,x_j) represent and pass through A matrixes X after transformation_i,x_jThe distance between.

6. according to the method described in claim 1, it is characterized in that,

The Local Metric matrix of the target part cluster learnt according to COS-LMNN algorithms, uses cosine similarity Algorithm calculates the input sample and at a distance from sample, forms second distance set in the cluster of the target part, including：

According to the Local Metric matrix for the target part cluster that COS-LMNN algorithms learn, cosine similarity algorithm is used Formula calculates the input sample and at a distance from sample, forms second distance set in the sample set；

Wherein, the cosine similarity algorithmic formula is：

The second distance set D2 includes：{D(x_i,x_s1), D (x_i,x_s2) ..., D (x_i,x_si)}；

Wherein, i represents the label of input sample, x_iI-th of input sample is represented as x；x_siIt represents and sample generic i；Office Portion's metric matrix is A_S；M_S=A_S ^TA_S；I takes positive integer；D(x_i,x_si) represent the input sample x under Local Metric matrix_iWith it is described In the cluster of target part at a distance from the sample generic with i；I takes positive integer；A_s(x_i,x_si) represent and pass through A_SX after matrixing_i,x_si The distance between.

7. a kind of cardiovascular and cerebrovascular disease risk profile device, which is characterized in that described device includes：

Gather acquisition module, for obtaining sample set；

Determined by multiple samples of the sample set according to the patient medical data library collection for setting up label；One sample packet It includes：Number, feature and the characteristic of patient；The label includes：First label and the second label；First tag identifier heart and brain Vascular disease's sample；The non-Patients with Cardiovascular/Cerebrovascular Diseases sample of second tag identifier；

Sample acquisition module, for obtaining an input sample；

The input sample is made of the medical treatment ＆ health physical examination data of patient to be predicted and the medical data merging of medical treatment；

Matrix computing module obtains described for carrying out metric learning using cosine-large-spacing nearest-neighbors COS-LMNN algorithms The global metric matrix of sample set；

Sample in sample set is divided into preset quantity by First partial cluster determining module for using preset clustering algorithm Local cluster；

First apart from determining module, for calculating the input using cosine similarity algorithm according to the global metric matrix Sample forms the first distance set in the sample set at a distance from sample；

First sample determining module, for according to preset first K values and first distance set, using k nearest neighbor algorithms, meter The first K values first for obtaining input sample are calculated adjacent to sample；

Second local cluster determining module, for determining described first adjacent to the local cluster where sample；

Target part cluster determining module, in the local cluster where sample, selection first to be adjacent to sample described first Quantity be more than the first predetermined threshold value local cluster, as target part cluster；

Local cluster division module, for the input sample to be included in target part cluster；

Second distance determining module, the Local Metric square of the target part cluster for being learnt according to COS-LMNN algorithms Battle array, is calculated using cosine similarity algorithm, the input sample in the cluster of the target part at a distance from sample, form second away from From set；

Second sample determining module is used in the cluster of the target part, according to preset 2nd K values and the second distance collection It closes, using k nearest neighbor algorithms, determines the 2nd K values second of the input sample adjacent to sample；

Statistical module, for counting the second the first label number and the second label number adjacent to sample；

Label determining module, if it is more than pre- to be used for second adjacent to the first label number of sample and the ratio of the second label number Bidding label threshold value, then using the first label as the label of input sample, otherwise using the second label as the label of input sample；

Clinical samples determining module determines whether input sample is cardiovascular and cerebrovascular disease for the label according to the input sample The sample of patient；

Patients with Cardiovascular/Cerebrovascular Diseases determining module, if being the sample of Patients with Cardiovascular/Cerebrovascular Diseases for input sample, it is determined that Patient to be predicted in input sample is Patients with Cardiovascular/Cerebrovascular Diseases；

Non- Patients with Cardiovascular/Cerebrovascular Diseases determining module, if for input sample not being the sample of Patients with Cardiovascular/Cerebrovascular Diseases, Determine that the patient to be predicted in input sample is not Patients with Cardiovascular/Cerebrovascular Diseases.

8. device according to claim 7, which is characterized in that described device further includes：

High-risk determining module, for determining whether the patient to be predicted is high-risk heart and brain blood according to the health follow-up data of patient Pipe Disease；

Suggestion module in hospital, if being high-risk Patients with Cardiovascular/Cerebrovascular Diseases for the patient to be predicted, to described to be predicted Patient makees the suggestion of hospitalization；

Increase physical examination suggestion module, if not being high-risk Patients with Cardiovascular/Cerebrovascular Diseases for the patient to be predicted, to described Patient to be predicted makees to increase the suggestion of the physical examination frequency；

Healthy determining module, for determining whether the patient to be predicted is healthy user according to the health follow-up data of patient；

Normal physical examination suggestion module keeps the normal patient if being healthy user for the patient to be predicted The suggestion of the normal physical examination frequency；

Patient's determining module is failed to pinpoint a disease in diagnosis, if not being healthy user for the patient to be predicted, the patient to be predicted is marked It is denoted as and fails to pinpoint a disease in diagnosis patient, patient medical data library collection is added in the characteristic for failing to pinpoint a disease in diagnosis patient；

Wherein, it is Patients with Cardiovascular/Cerebrovascular Diseases to fail to pinpoint a disease in diagnosis patient.

9. device according to claim 7, which is characterized in that the set acquisition module, including：

Sample deletes submodule, for multiple samples according to the patient medical data library collection that label is arranged, by sample missing values Sample more than first threshold makees sample delete processing；

The sample missing values are：The ratio of the feature quantity lacked in one sample and feature total quantity in the sample；

Feature deletes submodule, and for being searched in a plurality of sample after delete processing, feature missing values are more than second threshold Feature makees feature delete processing；

The feature missing values are：In the same feature of a plurality of sample, the feature quantity for the data that lack in individuality and same feature are total The ratio of quantity；

Fisrt feature submodule searches the feature of missing characteristic for a plurality of sample after making feature delete processing, makees For fisrt feature；

Missing Data Filling submodule makees missing values for using Multiple Imputation to the characteristic of fisrt feature missing It fills up；

Data classification submodule, for according to data type, the characteristic of a plurality of sample after being filled up to missing values to be done Classification obtains classification results；

Wherein, the classification results include：Discrete features data and continuous characteristic；

Data processing submodule, for according to classification results, by the discrete features data and continuous characteristic, work and data The corresponding processing of type；

Set update submodule, for the discrete features data and continuous characteristic to be done corresponding treated feature Patient medical data library collection is added in data, as first database collection；

Wherein, by the discrete features data and continuous characteristic, make processing corresponding with data type, including：To discrete Characteristic carries out one-hot coding；To continuous characteristic, it is standardized using just too standardization z-score methods；

Equilibrium treatment submodule, for carrying out uneven lack sampling and SMOTE algorithms to the sample of the first database collection Weighing apparatus processing, obtains the second data base set；

Variance deletes submodule, the side of the same characteristic for calculating second centralized database using method of analysis of variance Difference deletes the characteristic that characteristic variance yields is less than default variance threshold values, obtains third data base set；

Weight calculation submodule, the power for calculating each characteristic of third centralized database using relief algorithms Weight；

Score deletes submodule, for according to the weight of characteristic and the corresponding fractional value of the weight of characteristic, by third Centralized database fractional value is less than the characteristic of preset fraction threshold value and corresponding feature is deleted, and obtains the 4th data base set；

Gather determination sub-module, sample set is determined using forward selection procedures according to the 4th data base set.

10. a kind of electronic equipment, which is characterized in that including processor, communication interface, memory and communication bus, wherein processing Device, communication interface, memory complete mutual communication by communication bus；

Memory, for storing computer program；

Processor when for executing the program stored on memory, realizes any method and steps of claim 1-6.