CN111785384A - Artificial intelligence-based abnormal data identification method and related equipment - Google Patents

Abstract

The invention relates to the technical field of artificial intelligence, and provides an artificial intelligence-based abnormal data identification method, comprising: receiving input medical data; performing medical processing on the medical data to obtain the processing data; inputting the processing data into a pre-training In a good combined classifier model, a classification prediction result is obtained, wherein the combined classification model is obtained by training based on multiple dimensions; according to the classification prediction result, an abnormality judgment is performed on each cluster of the medical data. , to obtain the clustering abnormal result; by guiding the aggregation algorithm bagging, according to the clustering abnormal result, the abnormality judgment is performed on the medical treatment record matched by the medical data, and the abnormal medical treatment record result is obtained. The invention also relates to blockchain technology, which can upload abnormal results of medical records to the blockchain. The present application can be applied in smart medical scenarios, thereby promoting the construction of smart cities.

Description

Artificial intelligence-based abnormal data identification method and related equipment

technical field

The invention relates to the technical field of artificial intelligence, in particular to an artificial intelligence-based abnormal data identification method and related equipment.

Background technique

With the continuous improvement of the social medical security system, the problem of people's medical treatment has been solved. At present, after seeing a doctor, people can use the case data to reimburse part of the medical expenses, which greatly reduces the problem of people's medical expenses.

However, in practice, it is found that some illegal users will fabricate medical data to reimburse medical expenses. If the amount of medical expenses reimbursed by these illegal users is large, there will be insufficient funds to guarantee the medical reimbursement of legal users, which will undoubtedly affect the legal the legitimate rights and interests of users.

Therefore, how to identify abnormal risks in medical data is an urgent technical problem to be solved.

SUMMARY OF THE INVENTION

In view of the above, it is necessary to provide an abnormal data identification method and related equipment based on artificial intelligence, which can identify abnormal risks in medical data.

A first aspect of the present invention provides a method for identifying abnormal data based on artificial intelligence, and the method for identifying abnormal data based on artificial intelligence includes:

receive incoming medical data;

Perform medical processing on the medical data to obtain processed data;

Inputting the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions;

According to the classification prediction result, abnormality determination is performed on each cluster of the medical data to obtain an abnormal clustering result;

By guiding the clustering algorithm bagging, according to the clustering abnormal result, an abnormality judgment is performed on the medical treatment record matched with the medical data, and an abnormal medical treatment record result is obtained.

In a possible implementation manner, the performing medical processing on the medical data, and obtaining the processing data includes:

standard conversion is performed on the non-standard data in the medical data to obtain the first data;

Correcting erroneous data in the medical data to obtain second data;

Converting the age data in the medical data according to age to obtain third data;

Unprocessed data among the first data, the second data, the third data, and the medical data is determined as processed data.

In a possible implementation manner, performing an abnormality determination on each cluster of the medical data according to the classification prediction result, and obtaining the clustering abnormality result includes:

For each of the clusters, obtain an abnormal number threshold of the clusters;

Judging whether the number of processed data of the cluster as abnormal data in the classification prediction result exceeds the abnormal number threshold;

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data exceeds the abnormal number threshold, the cluster is determined to be an abnormal cluster; or

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data does not exceed the abnormal number threshold, it is determined that the cluster is a normal cluster.

In a possible implementation manner, by guiding the aggregation algorithm bagging, according to the clustering abnormal results, the abnormality judgment is performed on the medical treatment records matched by the medical data, and the abnormal results of the medical treatment records obtained include:

Divide the medical data according to the medical records, and obtain a plurality of medical records;

For each piece of medical treatment record data, by guiding the aggregation algorithm bagging, determine the first medical treatment record data with abnormal clustering and the second medical treatment record data without abnormal clustering;

An abnormality flag is respectively set for the first medical visit record data and the second medical visit record data.

In a possible implementation manner, before the receiving the input medical data, the artificial intelligence-based abnormal data identification method further includes:

Obtain the first test set, the second test set and the third test set;

Based on the Gaussian mixture model GMM, the first test set is trained to obtain a first classifier;

Based on the KL divergence, the second test set is trained to obtain a second classifier;

Based on the deep adversarial clustering DASC, the third test set is trained to obtain a third classifier;

The first classifier, the second classifier and the third classifier are combined to obtain a combined classifier model.

In a possible implementation, the artificial intelligence-based method for identifying abnormal data further includes:

According to the abnormal result of the medical treatment record, determine the abnormal medical treatment record data from a plurality of the medical treatment record data;

For the abnormal medical treatment record data, calculate the abnormal score of each abnormal medical treatment record data according to a preset clustering weight;

According to the abnormal score, the abnormality level of each abnormal medical treatment record data is determined.

In a possible implementation, the artificial intelligence-based method for identifying abnormal data further includes:

judging whether the abnormality level is higher than a preset abnormality level threshold;

If the abnormality level is higher than the preset abnormality level threshold, it is determined that the user to which the abnormal medical treatment record data belongs is a high-risk user.

A second aspect of the present invention provides an abnormal data identification device, characterized in that, the abnormal data identification device includes:

a receiving module for receiving the input medical data;

a processing module for performing medical processing on the medical data to obtain processing data;

an input module for inputting the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions;

A determination module, configured to perform an abnormal determination on each cluster of the medical data according to the classification prediction result, and obtain an abnormal cluster result;

The matching module is used for carrying out an abnormal judgment on the medical treatment record matched by the medical data according to the clustering abnormal result by bagging the guided aggregation algorithm, and obtaining the abnormal medical treatment record result;

The uploading module is used to upload the abnormal result of the medical treatment record to the blockchain.

A third aspect of the present invention provides an electronic device, the electronic device includes a processor and a memory, and the processor is configured to implement the artificial intelligence-based abnormal data identification method when executing a computer program stored in the memory.

A fourth aspect of the present invention provides a computer-readable storage medium, where a computer program is stored thereon, and when the computer program is executed by a processor, the artificial intelligence-based abnormal data identification method is implemented.

In the above technical solution, the medical data can be classified and predicted based on the classifiers of multiple dimensions, and the abnormality of medicine, abnormality of examination, abnormality of patient, abnormality of doctor, abnormality of medical institution, etc. can be identified, and finally whether the record of the visit is normal or not, then is marked as an exception record. The application of the multi-dimensional anomaly identification model in medical insurance risk control can accurately identify the anomalies of medical data. At the same time, it is not strongly dependent on medical knowledge and medical background, which is conducive to identifying high-risk fraudulent insurance users.

Description of drawings

FIG. 1 is a flowchart of a preferred embodiment of an artificial intelligence-based abnormal data identification method disclosed in the present invention.

FIG. 2 is a functional block diagram of a preferred embodiment of an abnormal data identification device disclosed in the present invention.

FIG. 3 is a schematic structural diagram of an electronic device implementing a preferred embodiment of the artificial intelligence-based abnormal data identification method according to the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The terms "comprising" and "having" and any variations thereof in the description and claims of this application are intended to cover non-exclusive inclusion, eg, a process, method, system, product or product comprising a series of steps or units. The apparatus is not necessarily limited to those steps or units expressly listed, but may include other steps or units not expressly listed or inherent to the process, method, product or apparatus.

In addition, the technical solutions between the various embodiments can be combined with each other, but must be based on the realization by those of ordinary skill in the art. When the combination of technical solutions is contradictory or cannot be realized, it should be considered that the combination of such technical solutions does not exist. , is not within the scope of protection required by the present invention.

The electronic device is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field programmable gates Arrays (FPGA), Digital Signal Processors (DSP), embedded devices, etc. The electronic device may include, but is not limited to, any electronic product that can interact with the user through a keyboard, a mouse, a remote control, a touchpad or a voice-activated device, for example, a personal computer, a tablet computer, a smart phone, a personal computer, a Digital Assistant PDA etc.

Please refer to FIG. 1. FIG. 1 is a flowchart of a preferred embodiment of an artificial intelligence-based abnormal data identification method disclosed in the present invention. Wherein, according to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

S11. Receive the input medical data.

The medical data may include, but not limited to, medical settlement data such as patient visit record ID, patient ID, doctor ID, medical institution ID, gender, age, primary diagnosis, social security directory name, and social security directory unified code.

S12. Perform medical processing on the medical data to obtain processed data.

Specifically, performing medical processing on the medical data and obtaining the processed data includes:

standard conversion is performed on the non-standard data in the medical data to obtain the first data;

Correcting erroneous data in the medical data to obtain second data;

Converting the age data in the medical data according to age to obtain third data;

Unprocessed data among the first data, the second data, the third data, and the medical data is determined as processed data.

Among them, (1) some data in the medical data are non-standard, and it is necessary to convert the non-standard data in the medical data into standard data. For example, in the gender description, different description methods are used, including "male" and "female". ”, “F” and “M”, etc., are unified into one standard for description; (2) Correct the erroneous data in the medical data, for example, in the description of the diagnosis and treatment item information, the occurrence of “Blood cell analysis_Five classification additional charges/ item", "blood cell analysis (five categories)", "blood cell analysis (five categories plus)/time", etc., then corrected to "(new) blood cell analysis_five categories plus"; (3) The age of the patient is calculated according to Age groups are converted to infants, toddlers, teenagers, youth, middle-aged and elderly, etc.

After the medical data is processed, it is beneficial to the unified identification and classification of the subsequent combined classifier model, and helps to improve the accuracy of the classification.

S13. Input the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions.

The combined classifier may be composed of multiple classifiers, and each classifier can cluster the processed data, and some discrete data that is not clustered into a certain category will be It is considered to be abnormal data, and the data that can be clustered into a certain class is considered to be normal data.

Wherein, the classification prediction result includes the clustering result of each classifier in the combined classifier to the processed data and the abnormal data result. The clustering result is the type of clustering, including but not limited to drug category, examination category, patient category, doctor category, and medical institution category. The abnormal data result means that the data of a certain cluster is abnormal, such as abnormal drugs, abnormal examinations, abnormal patients, abnormal doctors, or abnormal medical institutions.

Optionally, before the step S11, the method further includes:

Obtain the first test set, the second test set and the third test set;

Based on the Gaussian mixture model GMM, the first test set is trained to obtain a first classifier;

Based on the KL divergence, the second test set is trained to obtain a second classifier;

Based on the deep adversarial clustering DASC, the third test set is trained to obtain a third classifier;

The first classifier, the second classifier and the third classifier are combined to obtain a combined classifier model.

Among them, GMM refers to the estimation of the probability density distribution of the sample, and the training model used for estimation is the weighted sum of several Gaussian models. Each Gaussian model represents a cluster. The data in the sample is projected on several Gaussian models, and the probability of each class is obtained separately. Then, the class with the highest probability can be selected as the clustering result.

Assuming that the mixture Gaussian model consists of K Gaussian models (that is, the data contains K classes), the probability density function of GMM is as follows:

Among them, p(x|k)=N(x|u _k ,∑k) is the probability density function of the kth Gaussian model, which can be regarded as the probability that the model produces x after the kth model is selected, p( k)=π _k is the weight of the k-th Gaussian model, which is called the prior probability of selecting the k-th model, and satisfies

For example, the information describing a patient includes N-dimensional features such as age, gender, main diagnosis, prescribed drugs, and diagnosis and treatment. If there are already 2 patients, an N-dimensional two-component Gaussian mixture model is constructed at this time. According to the existing two patient data points, the parameter mean vector and covariance matrix of the Gaussian mixture model are determined, and the EM expectation maximization algorithm can be used to determine the parameters, and finally two GMMs describing the two patients are obtained. If a new patient comes, the new patient is represented by N-dimensional features and input into the calculated two Gaussian mixture models, and the probability under each GMM is obtained. Whichever probability is greater, then the new patient belongs to which category patient. By analogy, the clustering of patients is achieved.

where KL divergence is a measure of the asymmetry of the difference between two probability distributions P and Q. KL-divergence is used to measure the average number of bits per character used in this case. Can be used to measure the distance between two distributions, the formula is as follows:

Among them, P represents the true distribution of the data, and Q represents the theoretical distribution of the data, the model distribution, or the approximate distribution of P. If the optimal encoding of the probability distribution p(x) is used (that is, the encoding length of the character x is equal to log(1/P(x)) to encode the characters conforming to the distribution Q(x), then these characters will be represented than the ideal case. Use a few more bits.

For example, the information describing the doctor's diagnosis and treatment behavior and effect of diagnosis and treatment is as follows: doctor ID, ID of the medical institution where the doctor is located, number of visits, number of visits, proportion of expensive drugs, proportion of inspection costs, readmission rate, recurrence rate, return visit rate, etc. N-dimensional features. (1) If there are M doctor information, randomly select Q vectors from the M N-dimensional vectors as the initial mean clustering center, where Q is the number of types; (2) For each doctor information to be clustered, calculate the number of (3) assign a class number to the doctor to be clustered, and the class number is taken from the category of the mean center with the smallest distance from the fuzzy measure; (4) traverse the Q vectors, according to the Class number, calculate the average vector of vectors with the same class number respectively, and this average vector is updated as the new mean center; (5) For each category, calculate the fuzzy metric distance between the current mean center and the updated mean center; (6) ) If the fuzzy metric distance between the front and rear mean centers is smaller than the preset threshold, the classification ends, otherwise, go back to step (2). By analogy, the clustering of doctors is achieved.

Among them, the deep adversarial clustering (DASC) network consists of two parts: a subspace generator and a discriminator: in the generator, a convolutional autoencoder is used to learn the sample representation, and the self-expression layer is embedded in the encoder and decoder with To get the similarity matrix of the samples and use the Ncut method for clustering, and then generate "true samples" and "false samples" based on the clusters generated by the clustering; for the discriminator, receive the generated true and false samples, and distinguish the true samples and fake samples. Through DASC network training, the coefficient matrix CC is finally obtained, and the clustering result is obtained through spectral clustering.

Use deep adversarial subspace clustering to obtain the final clustering result, calculate the average value of patients in each category as the clustering center according to the clustering results, and then calculate the distance from each patient to the clustering center of this category, and select each The patients in the class closest to the cluster center get the patient subset, which is the result of patient clustering.

S14. According to the classification prediction result, perform abnormality determination on each cluster of the medical data to obtain a cluster abnormality result.

Specifically, according to the classification prediction result, the abnormality judgment is performed on each cluster of the medical data, and the clustering abnormal result obtained includes:

For each of the clusters, obtain an abnormal number threshold of the clusters;

Judging whether the number of processed data of the cluster as abnormal data in the classification prediction result exceeds the abnormal number threshold;

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data exceeds the abnormal number threshold, the cluster is determined to be an abnormal cluster; or

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data does not exceed the abnormal number threshold, it is determined that the cluster is a normal cluster.

For example, for checking this cluster, if more than 2 processed data in the classification prediction result are determined to be abnormal data, then the inspection cluster is considered to be an abnormal cluster. On the contrary, if the classification prediction result in Only 1 processing data is judged as abnormal data, then it is considered that the inspection cluster is not an abnormal cluster, that is, it belongs to a normal cluster.

S15 , performing an abnormality judgment on the medical treatment records matched by the medical data according to the clustering abnormal results by bagging the guided aggregation algorithm, and obtaining the abnormal results of the medical treatment records.

Specifically, according to the bagging of the guided aggregation algorithm, according to the abnormal clustering results, the abnormality judgment is performed on the medical records matched by the medical data, and the abnormal results of the medical records obtained include:

Divide the medical data according to the medical records, and obtain a plurality of medical records;

For each piece of medical treatment record data, by guiding the aggregation algorithm bagging, determine the first medical treatment record data with abnormal clustering and the second medical treatment record data without abnormal clustering;

An abnormality flag is respectively set for the first medical visit record data and the second medical visit record data.

Among them, each medical record data can include various clusters. If a certain cluster is abnormal in a medical record, the medical record can be considered as an abnormal record, and an abnormal flag can be set, such as "1". On the contrary, if a certain medical record is abnormal All clusters of the medical treatment record are normal, and the medical treatment record can be considered as a normal record, and a normal flag can be set, such as "0".

As shown in the table below:

The method also includes:

According to the abnormal result of the medical treatment record, determine the abnormal medical treatment record data from a plurality of the medical treatment record data;

For the abnormal medical treatment record data, calculate the abnormal score of each abnormal medical treatment record data according to a preset clustering weight;

According to the abnormal score, the abnormality level of each abnormal medical treatment record data is determined.

Among them, the clustering weight of each cluster can be set in advance, for example, clustering weight A is set for drugs, clustering weight B is set for inspection, clustering weight C is set for patients, clustering weight D is set for doctors, and clustering weight E is set for medical institutions , among which, the clustering weights of different clusters are different. The level of the clustering weights is mainly related to the importance of the clusters. The more important the clusters are, the higher the clustering weights are. For example, the clustering weights of medical institutions are higher than Drugs have a high clustering weight.

Among them, after calculating the abnormal score of each abnormal medical treatment record data, the abnormal degree level can be set according to the abnormal score. Specifically, the abnormal score threshold can be set, and the abnormal score can be compared with the abnormal score threshold. Comparison to determine the level of abnormality, which can be divided into severe abnormality, moderate abnormality and easy abnormality.

Optionally, the method further includes:

judging whether the abnormality level is higher than a preset abnormality level threshold;

If the abnormality level is higher than the preset abnormality level threshold, it is determined that the user to which the abnormal medical treatment record data belongs is a high-risk user.

Wherein, if the abnormality level is higher than the preset abnormality level threshold, it indicates that there is a very serious abnormality in the abnormal medical treatment record data, and to a certain extent, it also reflects that the abnormality may be artificially fabricated false data. The user to which the abnormal medical treatment record data belongs may also be a high-risk user.

Through the above methods, high-risk fraudsters can be identified, which is beneficial to the management of medical insurance risk control.

Optionally, the method further includes:

Upload the abnormal result of the medical treatment record to the blockchain.

Among them, in order to ensure the privacy and security of the data, it is necessary to upload the abnormal results of the medical treatment records to the blockchain for storage.

In the method flow described in Fig. 1, the medical data can be classified and predicted based on the classifiers of multiple dimensions, and the abnormality of medicine, abnormality of examination, abnormality of patient, abnormality of doctor, abnormality of medical institution, etc. can be identified, and finally the record of this visit can be judged. If it is normal, it will be marked as an abnormal record. The application of the multi-dimensional anomaly identification model in medical insurance risk control can accurately identify the anomalies of medical data. At the same time, it is not strongly dependent on medical knowledge and medical background, which is conducive to identifying high-risk fraudulent insurance users.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. improvements, but these all belong to the protection scope of the present invention.

Please refer to FIG. 2 , which is a functional block diagram of a preferred embodiment of an abnormal data identification device disclosed in the present invention.

In some embodiments, the abnormal data identification apparatus is executed in an electronic device. The abnormal data identification device may include a plurality of functional modules composed of program code segments. The program codes of each program segment in the abnormal data identification device can be stored in the memory and executed by at least one processor to execute part or all of the steps in the artificial intelligence-based abnormal data identification method described in FIG. 1 . , for details, reference may be made to the relevant description in FIG. 1 , which will not be repeated here.

In this embodiment, the abnormal data identification device may be divided into a plurality of functional modules according to the functions performed by the device. The functional modules may include: a receiving module 201 , a processing module 202 , an input module 203 , a determining module 204 and a matching module 205 . The modules referred to in the present invention refer to a series of computer program segments that can be executed by at least one processor and can perform fixed functions, and are stored in a memory. In some embodiments, the functions of each module will be described in detail in this embodiment.

The receiving module 201 is used for receiving input medical data.

The medical data may include, but not limited to, medical settlement data such as patient visit record ID, patient ID, doctor ID, medical institution ID, gender, age, primary diagnosis, social security directory name, and social security directory unified code.

The processing module 202 is configured to perform medical processing on the medical data to obtain processing data.

Specifically, performing medical processing on the medical data and obtaining the processed data includes:

standard conversion is performed on the non-standard data in the medical data to obtain the first data;

Correcting erroneous data in the medical data to obtain second data;

Converting the age data in the medical data according to age to obtain third data;

Unprocessed data among the first data, the second data, the third data, and the medical data is determined as processed data.

Among them, (1) some data in the medical data are non-standard, and it is necessary to convert the non-standard data in the medical data into standard data. For example, in the gender description, different description methods are used, including "male" and "female". ”, “F” and “M”, etc., are unified into one standard for description; (2) Correct the erroneous data in the medical data, for example, in the description of the diagnosis and treatment item information, the occurrence of “Blood cell analysis_Five classification additional charges/ item", "blood cell analysis (five categories)", "blood cell analysis (five categories plus)/time", etc., then corrected to "(new) blood cell analysis_five categories plus"; (3) The age of the patient is calculated according to Age groups are converted to infants, toddlers, teenagers, youth, middle-aged and elderly, etc.

After the medical data is processed, it is beneficial to the unified identification and classification of the subsequent combined classifier model, and helps to improve the accuracy of the classification.

The input module 203 is configured to input the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions.

The combined classifier may be composed of multiple classifiers, and each classifier can cluster the processed data, and some discrete data that is not clustered into a certain category will be It is considered to be abnormal data, and the data that can be clustered into a certain class is considered to be normal data.

Wherein, the classification prediction result includes the clustering result of each classifier in the combined classifier to the processed data and the abnormal data result. The clustering result is the type of clustering, including but not limited to drug category, examination category, patient category, doctor category, and medical institution category. The abnormal data result means that the data of a certain cluster is abnormal, such as abnormal drugs, abnormal examinations, abnormal patients, abnormal doctors, or abnormal medical institutions.

The determination module 204 is configured to perform abnormality determination on each cluster of the medical data according to the classification prediction result, and obtain a cluster abnormality result.

Specifically, according to the classification prediction result, the abnormality judgment is performed on each cluster of the medical data, and the clustering abnormal result obtained includes:

For each of the clusters, obtain an abnormal number threshold of the clusters;

Judging whether the number of processed data of the cluster as abnormal data in the classification prediction result exceeds the abnormal number threshold;

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data exceeds the abnormal number threshold, the cluster is determined to be an abnormal cluster; or

If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data does not exceed the abnormal number threshold, it is determined that the cluster is a normal cluster.

For example, for checking this cluster, if more than 2 processed data in the classification prediction result are determined to be abnormal data, then the inspection cluster is considered to be an abnormal cluster. On the contrary, if the classification prediction result in Only 1 processing data is judged as abnormal data, then it is considered that the inspection cluster is not an abnormal cluster, that is, it belongs to a normal cluster.

The matching module 205 is configured to perform an abnormality judgment on the medical treatment records matched by the medical data according to the clustering abnormal results through bagging guided aggregation algorithm, and obtain the abnormal medical treatment records results.

Specifically, according to the bagging of the guided aggregation algorithm, according to the abnormal clustering results, the abnormality judgment is performed on the medical records matched by the medical data, and the abnormal results of the medical records obtained include:

Divide the medical data according to the medical records, and obtain a plurality of medical records;

For each piece of medical treatment record data, by guiding the aggregation algorithm bagging, determine the first medical treatment record data with abnormal clustering and the second medical treatment record data without abnormal clustering;

An abnormality flag is respectively set for the first medical visit record data and the second medical visit record data.

Among them, each medical record data can include various clusters. If a certain cluster is abnormal in a medical record, the medical record can be considered as an abnormal record, and an abnormal flag can be set, such as "1". On the contrary, if a certain medical record is abnormal All clusters of the medical treatment record are normal, and the medical treatment record can be considered as a normal record, and a normal flag can be set, such as "0".

As shown in the table below:

Optionally, the abnormal data identification device further includes:

a determining module, configured to determine abnormal medical treatment record data from a plurality of the medical treatment record data according to the abnormal result of the medical treatment record;

a calculation module, configured to calculate the abnormal score of each abnormal medical visit record data according to a preset clustering weight for the abnormal medical visit record data;

The determining module is further configured to determine, according to the abnormal score, the abnormality level of each of the abnormal medical treatment record data.

Among them, the clustering weight of each cluster can be set in advance, for example, clustering weight A is set for drugs, clustering weight B is set for inspection, clustering weight C is set for patients, clustering weight D is set for doctors, and clustering weight E is set for medical institutions , among which, the clustering weights of different clusters are different. The level of the clustering weights is mainly related to the importance of the clusters. The more important the clusters are, the higher the clustering weights are. For example, the clustering weights of medical institutions are higher than Drugs have a high clustering weight.

Among them, after calculating the abnormal score of each abnormal medical treatment record data, the abnormal degree level can be set according to the abnormal score. Specifically, the abnormal score threshold can be set, and the abnormal score can be compared with the abnormal score threshold. Comparison to determine the level of abnormality, which can be divided into severe abnormality, moderate abnormality and easy abnormality.

Optionally, the determination module is further configured to determine whether the abnormality level is higher than a preset abnormality level threshold;

The determining module is further configured to determine that the user to which the abnormal medical treatment record data belongs is a high-risk user if the abnormality degree level is higher than a preset abnormality level threshold.

Wherein, if the abnormality level is higher than the preset abnormality level threshold, it indicates that there is a very serious abnormality in the abnormal medical treatment record data, and to a certain extent, it also reflects that the abnormality may be artificially fabricated false data. The user to which the abnormal medical treatment record data belongs may also be a high-risk user.

Through the above methods, high-risk fraudsters can be identified, which is beneficial to the management of medical insurance risk control.

Optionally, the abnormal data identification device further includes:

The uploading module is used to upload the abnormal result of the medical treatment record to the blockchain.

Optionally, the abnormal data identification device further includes:

an acquisition module for acquiring the first test set, the second test set and the third test set;

a training module for training the first test set based on the Gaussian mixture model GMM to obtain a first classifier;

The training module is further configured to perform training on the second test set based on the KL divergence to obtain a second classifier;

The training module is also used to train the third test set based on the deep confrontation clustering DASC to obtain a third classifier;

The combination module is configured to combine the first classifier, the second classifier and the third classifier to obtain a combined classifier model.

Among them, GMM refers to the estimation of the probability density distribution of the sample, and the training model used for estimation is the weighted sum of several Gaussian models. Each Gaussian model represents a cluster. The data in the sample is projected on several Gaussian models, and the probability of each class is obtained separately. Then, the class with the highest probability can be selected as the clustering result.

Assuming that the mixture Gaussian model consists of K Gaussian models (that is, the data contains K classes), the probability density function of GMM is as follows:

Among them, p(x|k)=N(x|u _k ,∑k) is the probability density function of the kth Gaussian model, which can be regarded as the probability that the model produces x after the kth model is selected, p( k)=π _k is the weight of the k-th Gaussian model, which is called the prior probability of selecting the k-th model, and satisfies

For example, the information describing a patient includes N-dimensional features such as age, gender, main diagnosis, prescribed drugs, and diagnosis and treatment. If there are already 2 patients, an N-dimensional two-component Gaussian mixture model is constructed at this time. According to the existing two patient data points, the parameter mean vector and covariance matrix of the Gaussian mixture model are determined, and the EM expectation maximization algorithm can be used to determine the parameters, and finally two GMMs describing the two patients are obtained. If a new patient comes, the new patient is represented by N-dimensional features and input into the calculated two Gaussian mixture models, and the probability under each GMM is obtained. Whichever probability is greater, then the new patient belongs to which category patient. By analogy, the clustering of patients is achieved.

where KL divergence is a measure of the asymmetry of the difference between two probability distributions P and Q. KL-divergence is used to measure the average number of bits per character used in this case. Can be used to measure the distance between two distributions, the formula is as follows:

Among them, P represents the true distribution of the data, and Q represents the theoretical distribution of the data, the model distribution, or the approximate distribution of P. If the optimal encoding of the probability distribution p(x) is used (that is, the encoding length of the character x is equal to log(1/P(x)) to encode the characters conforming to the distribution Q(x), then these characters will be represented than the ideal case. Use a few more bits.

For example, the information describing the doctor's diagnosis and treatment behavior and effect of diagnosis and treatment is as follows: doctor ID, ID of the medical institution where the doctor is located, number of visits, number of visits, proportion of expensive drugs, proportion of inspection costs, readmission rate, recurrence rate, return visit rate, etc. N-dimensional features. (1) If there are M doctor information, randomly select Q vectors from the M N-dimensional vectors as the initial mean clustering center, where Q is the number of types; (2) For each doctor information to be clustered, calculate the number of (3) assign a class number to the doctor to be clustered, and the class number is taken from the category of the mean center with the smallest distance from the fuzzy measure; (4) traverse the Q vectors, according to the Class number, calculate the average vector of vectors with the same class number respectively, and this average vector is updated as the new mean center; (5) For each category, calculate the fuzzy metric distance between the current mean center and the updated mean center; (6) ) If the fuzzy metric distance between the front and rear mean centers is smaller than the preset threshold, the classification ends, otherwise, go back to step (2). By analogy, the clustering of doctors is achieved.

Among them, the deep adversarial clustering (DASC) network consists of two parts: a subspace generator and a discriminator: in the generator, a convolutional autoencoder is used to learn the sample representation, and the self-expression layer is embedded in the encoder and decoder with To get the similarity matrix of the samples and use the Ncut method for clustering, and then generate "true samples" and "false samples" based on the clusters generated by the clustering; for the discriminator, receive the generated true and false samples, and distinguish the true samples and fake samples. Through DASC network training, the coefficient matrix CC is finally obtained, and the clustering result is obtained through spectral clustering.

Use deep adversarial subspace clustering to obtain the final clustering result, calculate the average value of patients in each category as the clustering center according to the clustering results, and then calculate the distance from each patient to the clustering center of this category, and select each The patients in the class closest to the cluster center get the patient subset, which is the result of patient clustering.

In the abnormal data identification device described in FIG. 2 , the medical data can be classified and predicted based on the classifiers of multiple dimensions, and the abnormal drug, abnormal examination, abnormal patient, abnormal doctor, abnormal medical institution, etc. can be identified, and the final judgment of the time Whether the medical record is normal, it is marked as an abnormal record. The application of the multi-dimensional anomaly identification model in medical insurance risk control can accurately identify the anomalies of medical data. At the same time, it is not strongly dependent on medical knowledge and medical background, which is conducive to identifying high-risk fraudulent insurance users.

As shown in FIG. 3 , FIG. 3 is a schematic structural diagram of an electronic device implementing a preferred embodiment of the artificial intelligence-based abnormal data identification method of the present invention. The electronic device 3 includes a memory 31 , at least one processor 32 , a computer program 33 stored in the memory 31 and executable on the at least one processor 32 , and at least one communication bus 34 .

Those skilled in the art can understand that the schematic diagram shown in FIG. 3 is only an example of the electronic device 3, and does not constitute a limitation on the electronic device 3, and may include more or less components than those shown, or combinations thereof Certain components, or different components, for example, the electronic device 3 may also include input and output devices, network access devices, and the like.

The at least one processor 32 may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC) ), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 32 can be a microprocessor or the processor 32 can also be any conventional processor, etc. The processor 32 is the control center of the electronic device 3, and uses various interfaces and lines to connect the entire electronic device 3 of each part.

The memory 31 can be used to store the computer program 33 and/or modules/units, and the processor 32 executes or executes the computer programs and/or modules/units stored in the memory 31 and calls the computer programs and/or modules/units stored in the memory 31. 31 to realize various functions of the electronic device 3 . The memory 31 may mainly include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may Data such as audio data and the like created in accordance with the use of the electronic device 3 are stored. In addition, the memory 31 may include non-volatile memory, such as hard disk, internal memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card (Flash Card), At least one disk storage device, flash memory device, or other non-volatile solid state storage device.

1, the memory 31 in the electronic device 3 stores a plurality of instructions to implement a method for identifying abnormal data based on artificial intelligence, and the processor 32 can execute the plurality of instructions to achieve:

receive incoming medical data;

Perform medical processing on the medical data to obtain processed data;

Inputting the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions;

According to the classification prediction result, abnormality determination is performed on each cluster of the medical data to obtain an abnormal clustering result;

By guiding the clustering algorithm bagging, according to the clustering abnormal results, the abnormality judgment is performed on the medical treatment records matched by the medical data, and the abnormal results of the medical treatment records are obtained;

Upload the abnormal result of the medical treatment record to the blockchain.

Specifically, for the specific implementation method of the above-mentioned instruction by the processor 32, reference may be made to the description of the relevant steps in the corresponding embodiment of FIG. 1, and details are not described herein.

In the electronic device 3 described in FIG. 3 , the medical data can be classified and predicted based on the classifiers of multiple dimensions, and the abnormality of medicine, abnormality of examination, abnormality of patient, abnormality of doctor, abnormality of medical institution, etc. can be identified, and finally the visit to the doctor can be judged. If the record is normal, it will be marked as abnormal record. The application of the multi-dimensional anomaly identification model in medical insurance risk control can accurately identify the anomalies of medical data. At the same time, it is not strongly dependent on medical knowledge and medical background, which is conducive to identifying high-risk fraudulent insurance users.

If the modules/units integrated in the electronic device 3 are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, and can also be completed by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments can be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form, and the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a removable hard disk, a magnetic disk, an optical disc, a computer memory, and a read-only memory (ROM, Read-Only Memory). .

In the several embodiments provided by the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division manners in actual implementation.

The modules described as separate components may or may not be physically separated, and components shown as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of hardware plus software function modules.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, but that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the invention is to be defined by the appended claims rather than the foregoing description, which are therefore intended to fall within the scope of the claims. All changes within the meaning and range of the equivalents of , are included in the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim. Several units or means recited in the system claims can also be realized by software or hardware.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be Modifications or equivalent substitutions can be made without departing from the spirit and scope of the technical solutions of the present invention.

Claims (10)

1. a method for identifying abnormal data based on artificial intelligence, is characterized in that, the method for identifying abnormal data based on artificial intelligence comprises: receive incoming medical data; Perform medical processing on the medical data to obtain processed data; Inputting the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions; According to the classification prediction result, abnormality determination is performed on each cluster of the medical data to obtain an abnormal clustering result; By guiding the clustering algorithm bagging, according to the clustering abnormal result, an abnormality judgment is performed on the medical treatment record matched with the medical data, and an abnormal medical treatment record result is obtained. 2. The method for identifying abnormal data based on artificial intelligence according to claim 1, characterized in that, performing medical processing on the medical data and obtaining the processed data comprises: standard conversion is performed on the non-standard data in the medical data to obtain the first data; Correcting erroneous data in the medical data to obtain second data; Converting the age data in the medical data according to age to obtain third data; Unprocessed data among the first data, the second data, the third data, and the medical data is determined as processed data. 3. The method for identifying abnormal data based on artificial intelligence according to claim 1, characterized in that, according to the classification prediction result, abnormal judgment is performed on each cluster of the medical data to obtain a cluster abnormal result. include: For each of the clusters, obtain an abnormal number threshold of the clusters; Judging whether the number of processed data of the cluster as abnormal data in the classification prediction result exceeds the abnormal number threshold; If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data exceeds the abnormal number threshold, the cluster is determined to be an abnormal cluster; or If, in the classification prediction result, it is determined that the number of processed data of the cluster as abnormal data does not exceed the abnormal number threshold, it is determined that the cluster is a normal cluster. 4. The method for identifying abnormal data based on artificial intelligence according to claim 1, characterized in that, according to the clustering abnormal result, an abnormality is performed on the medical treatment record matched by the medical data by bagging a guided aggregation algorithm. It is determined that the abnormal results obtained from the medical records include: Divide the medical data according to the medical records, and obtain a plurality of medical records; For each piece of medical treatment record data, by guiding the aggregation algorithm bagging, determine the first medical treatment record data with abnormal clustering and the second medical treatment record data without abnormal clustering; An abnormality flag is respectively set for the first medical visit record data and the second medical visit record data. 5. The artificial intelligence-based abnormal data identification method according to any one of claims 1 to 4, characterized in that, before the described receiving input medical data, the artificial intelligence-based abnormal data identification method further comprises: Obtain the first test set, the second test set and the third test set; Based on the Gaussian mixture model GMM, the first test set is trained to obtain a first classifier; Based on the KL divergence, the second test set is trained to obtain a second classifier; Based on the deep adversarial clustering DASC, the third test set is trained to obtain a third classifier; The first classifier, the second classifier and the third classifier are combined to obtain a combined classifier model. 6. The method for identifying abnormal data based on artificial intelligence according to any one of claims 1 to 4, wherein the method for identifying abnormal data based on artificial intelligence also comprises: According to the abnormal result of the medical treatment record, determine the abnormal medical treatment record data from a plurality of the medical treatment record data; For the abnormal medical treatment record data, calculate the abnormal score of each abnormal medical treatment record data according to a preset clustering weight; According to the abnormal score, the abnormality level of each abnormal medical treatment record data is determined. 7. The method for identifying abnormal data based on artificial intelligence according to any one of claims 1 to 4, wherein the method for identifying abnormal data based on artificial intelligence also comprises: judging whether the abnormality level is higher than a preset abnormality level threshold; If the abnormality level is higher than the preset abnormality level threshold, it is determined that the user to which the abnormal medical treatment record data belongs is a high-risk user. 8. A device for identifying abnormal data, wherein the device for identifying abnormal data comprises: a receiving module for receiving the input medical data; a processing module for performing medical processing on the medical data to obtain processing data; an input module for inputting the processed data into a pre-trained combined classifier model to obtain a classification prediction result, wherein the combined classification model is obtained by training based on multiple dimensions; A determination module, configured to perform an abnormal determination on each cluster of the medical data according to the classification prediction result, and obtain an abnormal cluster result; The matching module is used for carrying out an abnormal judgment on the medical treatment record matched by the medical data according to the clustering abnormal result by bagging the guided aggregation algorithm, and obtaining the abnormal medical treatment record result; The uploading module is used to upload the abnormal result of the medical treatment record to the blockchain. 9. An electronic device, characterized in that the electronic device comprises a processor and a memory, and the processor is configured to execute a computer program stored in the memory to implement the manual-based method according to any one of claims 1 to 7. An intelligent method for identifying abnormal data. 10. A computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, implements the method according to any one of claims 1 to 7. An artificial intelligence-based method for identifying abnormal data.

Images