CN108052979A - The method, apparatus and equipment merged to model predication value - Google Patents

Abstract

Disclosed are a method, device and equipment for fusing model prediction values, wherein the method for fusing model prediction values includes: based on a number of given samples, according to the set binning method to respectively classify the prediction values of the online prediction model binning with the prediction value of the offline prediction model; according to the result of binning, the first prediction value of each sample is transformed into the first interval feature corresponding to the interval in which the first prediction value is located, and the second prediction value of each sample is The predicted value is converted into the second interval feature corresponding to the interval in which the second predicted value is located; the converted sample data is composed of the first interval feature, the second interval feature and the label of the sample corresponding to each sample , and use the converted sample data to train the model, and the trained model is used to fuse the prediction value of the online prediction model and the prediction value of the offline prediction model to obtain the final prediction value.

Description

Method, device and equipment for fusing model prediction values

technical field

This specification relates to the technical field of machine learning, and in particular to a method, device and equipment for fusing model prediction values.

Background technique

Machine learning algorithm is a kind of algorithm that can automatically analyze and obtain laws from data, and use the laws to predict unknown data. It is widely used in many fields.

In practical applications, it includes online prediction model and offline prediction model. Among them, the offline prediction model is usually implemented by timing tasks. Its advantage is that it can incorporate features with higher dimensions and use more complex algorithms to achieve more accurate predictions. effect; however, due to the large number of features and complex algorithms, the prediction process is usually time-consuming. Compared with the offline prediction model, the online prediction model can use lower-dimensional features and simpler algorithms to achieve more efficient prediction. The disadvantage is that the features are not rich enough and the accuracy is not high. It can be seen that the online prediction model and the offline prediction model have their own advantages, and how to integrate them reasonably is an urgent problem in the industry.

Contents of the invention

In view of the above technical problems, the embodiment of this specification provides a method, device and equipment for fusing model prediction values. The technical solution is as follows:

In one aspect, a method for fusing model predictions is proposed, including:

Based on several given samples, according to the set binning method, the predicted value of the online prediction model and the predicted value of the offline prediction model are binned respectively, wherein each sample in the several samples includes: a first predicted value , the second predicted value and the label of the sample, the first predicted value is predicted by the online prediction model, and the second predicted value is predicted by the offline prediction model;

According to the result of binning, the first predicted value of each sample is converted into the first interval feature corresponding to the interval in which the first predicted value is located, and the second predicted value of each sample is converted into a feature corresponding to the interval of the second predicted value. The second interval feature corresponding to the interval at ;

The transformed sample data is composed of the first interval feature, the second interval feature and the label of the sample corresponding to each sample, and the transformed sample data is used to train the model, and the trained model is used for online The prediction value of the prediction model and the prediction value of the offline prediction model are fused to obtain the final prediction value.

In one aspect, a method for fusing model predictions is proposed, including:

Obtaining business data generated by the target user within the first time period, determining input features according to the business data and inputting them into the online prediction model, and outputting a first predicted value;

Acquiring a second prediction value corresponding to the target user obtained by using an offline prediction model, wherein the input characteristics of the offline prediction model are determined according to service characteristics generated by the target user within a second time period;

Obtain the results of binning the first predicted value of the online prediction model and the second predicted value of the offline prediction model, and respectively determine the first interval where the first predicted value is located and the second interval where the second predicted value is located. Two intervals;

According to the first interval and the second interval, the model obtained in advance is used to fuse the first predicted value and the second predicted value to obtain a final fused predicted value, and the fused predicted value is used to determine the label of the target user.

In one aspect, a device for fusing model predictions is proposed, including:

The binning unit, based on several given samples, performs binning on the predicted values of the online prediction model and the predicted values of the offline prediction model according to the set binning method, wherein each sample in the several samples includes: a first predicted value, a second predicted value, and a label of the sample, the first predicted value is predicted by an online prediction model, and the second predicted value is predicted by an offline prediction model;

The feature conversion unit converts the first predicted value of each sample into a first interval feature corresponding to the interval in which the first predicted value is located according to the result of binning, and converts the second predicted value of each sample into a feature corresponding to the first interval of the first predicted value. The second interval feature corresponding to the interval where the second predicted value is located;

The training unit uses the first interval feature, the second interval feature and the label of the sample corresponding to each sample to form the converted sample data, and uses the converted sample data to train the model, and the trained model is used It is used to fuse the prediction value of the online prediction model and the prediction value of the offline prediction model to obtain the final prediction value.

In one aspect, a device for fusing model predictions is proposed, including:

The online score prediction unit obtains the business data generated by the target user in the first time period before the trigger moment, determines the input characteristics according to the business data and inputs them into the online prediction model, and outputs the first predicted value. The online prediction model Labels for predicting users;

The offline score obtaining unit is configured to obtain a second predicted value corresponding to the target user obtained by using an offline prediction model, wherein the input characteristics of the offline prediction model are generated according to the target user in the past second time period Determined by business characteristics, the offline prediction model is used to predict the label of the user;

The interval determination unit determines the first interval in which the first prediction value is located and the interval in which the second prediction value is located according to the results of pre-binning the prediction values of the online prediction model and the prediction values of the offline prediction model. second interval;

The score fusion unit, according to the first interval and the second interval, uses the pre-trained model to fuse the first predicted value and the second predicted value to obtain a final fused predicted value, so The fusion prediction value is used to determine the label of the target user.

In one aspect, a computer device is provided, comprising:

processor;

memory for storing processor-executable instructions;

The processor is configured to:

Based on several given samples, according to the set binning method, the predicted value of the online prediction model and the predicted value of the offline prediction model are binned respectively, wherein each sample in the several samples includes: a first predicted value , the second predicted value and the label of the sample, the first predicted value is predicted by the online prediction model, and the second predicted value is predicted by the offline prediction model;

According to the result of binning, the first predicted value of each sample is converted into the first interval feature corresponding to the interval in which the first predicted value is located, and the second predicted value of each sample is converted into a feature corresponding to the interval of the second predicted value. The second interval feature corresponding to the interval at ;

The transformed sample data is composed of the first interval feature, the second interval feature and the label of the sample corresponding to each sample, and the transformed sample data is used to train the model, and the trained model is used for online The prediction value of the prediction model and the prediction value of the offline prediction model are fused to obtain the final prediction value.

In one aspect, a computer device is provided, comprising:

processor;

memory for storing processor-executable instructions;

The processor is configured to:

The online score prediction unit obtains the business data generated by the target user in the first time period before the trigger moment, determines the input characteristics according to the business data and inputs them into the online prediction model, and outputs the first predicted value. The online prediction model Labels for predicting users;

The offline score obtaining unit is configured to obtain a second predicted value corresponding to the target user obtained by using an offline prediction model, wherein the input characteristics of the offline prediction model are generated according to the target user in the past second time period Determined by business characteristics, the offline prediction model is used to predict the label of the user;

The interval determination unit determines the first interval in which the first prediction value is located and the interval in which the second prediction value is located according to the results of pre-binning the prediction values of the online prediction model and the prediction values of the offline prediction model. second interval;

The score fusion unit, according to the first interval and the second interval, uses the pre-trained model to fuse the first predicted value and the second predicted value to obtain a final fused predicted value, so The fusion prediction value is used to determine the label of the target user.

The effects produced by the technical solutions provided in the embodiments of this specification include:

The prediction value of the line prediction model and the prediction value of the offline prediction model are fused using the model obtained by machine learning, and finally the score obtained by fusion is used to predict the user's label, thereby improving the user's While predicting the accuracy of tags, it also meets the business requirements for low latency.

It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and are not intended to limit the embodiments of this specification.

In addition, any embodiment in the embodiments of this specification does not need to achieve all the above-mentioned effects.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of this specification 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 drawings in the following description are only These are some embodiments described in the embodiments of this specification, and those skilled in the art can also obtain other drawings based on these drawings.

Fig. 1 is a schematic flowchart of a method for fusing model prediction values provided by an embodiment of this specification;

FIG. 2 is a process of determining fusion weights provided by an embodiment of this specification;

Fig. 3 is a schematic structural diagram of a device for fusing model prediction values (weight training stage) provided by an embodiment of this specification;

Fig. 4 is a schematic structural diagram of a device for fusing model prediction values (score fusion stage) provided by an embodiment of this specification;

Fig. 5 is a schematic structural diagram of a device for configuring the device of the embodiment of the present specification.

Detailed ways

In order for those skilled in the art to better understand the technical solutions in the embodiments of this specification, the technical solutions in the embodiments of this specification will be described in detail below in conjunction with the drawings in the embodiments of this specification. Obviously, the described implementation Examples are only some of the embodiments in this specification, not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in this specification shall fall within the scope of protection.

Referring to Figure 1, in one embodiment of this specification, a method for fusing model prediction values is used to fuse the scores obtained by the online prediction model and the scores obtained by the offline prediction model. The method may include the following steps 101-104, wherein:

Step 101: Obtain the business data generated by the target user within the first time period, determine the input features according to the business data and input them into the online prediction model, and output the first predicted value.

Step 102: Obtain the second predicted value corresponding to the target user obtained by using the offline prediction model, wherein the input characteristics of the offline prediction model are determined according to the business characteristics generated by the target user within the second time period of.

Herein, both the online prediction model and the offline prediction model are models constructed using machine learning algorithms and used to predict user tags. The user tags that these two models need to predict can be related to specific businesses. For example, for an online payment business, the user tags that need to be predicted can be divided into: "high-risk users", "medium-risk users", " low-risk users", and so on. For an information recommendation service, the user labels to be predicted can be classified into: "sports", "education", "finance", etc. Both the online prediction model and the offline prediction model are trained with a certain number of training samples, and each of these training samples may include: a sample generated by a sample user in the process of participating in a specific service (such as an online payment service) or a variety of behavioral data, as well as the identified labels of sample users. Wherein, the same batch of samples can be used to train the online prediction model and the offline prediction model, or two different batches of samples can be used to train the online prediction model and the offline prediction model, which is not limited herein.

In the embodiment of this specification, the offline prediction model can be implemented through timed tasks, such as: perform an offline score prediction at a specified time or a specified time period every day, and the prediction process can be aimed at all users; while online prediction The model can be triggered by a specific user's operation, for example, a user's click on a webpage can trigger a score calculation process of the online prediction model.

Because offline prediction models usually use higher-dimensional feature data than online prediction models, the time span of feature data can also be longer, and more complex algorithms can be used. As shown in Figure 1, as a specific example, on day T, the offline prediction model can obtain the business data (feature A) generated by each user in the process of participating in a specific business on day T-1, according to the obtained business data The data (feature A) is processed accordingly, and the input features can be obtained and input into the offline prediction model, and the offline prediction score of each user (that is, the second prediction value in the text) is obtained and written into the database X. For the online prediction model, the user's online characteristic data (feature B) can be continuously collected and written into the database Y, wherein the online characteristic data can be quasi-real-time business generated by the user in the process of participating in a specific business Data, for example: the triggering time of the online prediction is t1, then the online feature data may be the business data generated during the time period of t0-t1 (for example, 3 minutes). It can be seen that after the user request for initiating the prediction process arrives, the scheduler needs to do two tasks. One is to read the second prediction value corresponding to the target user obtained by the latest offline prediction model calculation from the database X; The second is to read the online feature data of the target user from the database Y to carry out the next process of score prediction of the online prediction model.

So far, for any target user, a prediction score can be obtained through the online prediction model, and a prediction score can be obtained through the offline prediction model.

Step 103: According to the results of pre-binning the predicted values of the online prediction model and the predicted values of the offline prediction model, respectively determine the first interval in which the first predicted value is located and the first interval in which the second predicted value is located. Two intervals.

Step 104: According to the first interval and the second interval, use the pre-trained model to fuse the first predicted value and the second predicted value to obtain a final fused predicted value, wherein the The fusion prediction value is used to determine the label of the target user.

In an optional embodiment, step 104 may specifically include:

Step 1041: Obtain a first weight corresponding to the first interval and a second weight corresponding to the second interval based on the predetermined weights corresponding to each interval obtained by binning. Wherein, the parameters to be trained of the model include weights corresponding to each interval obtained by binning.

Step 1042: Using the first weight and the second weight to determine a fusion prediction value, the fusion prediction value is used to determine the label of the target user.

Since the above steps 103 to 104 need to be implemented based on the binning results and the weights corresponding to the intervals obtained by binning, therefore, before introducing steps 103 to 104 in detail, it is necessary to introduce a method for determining fusion weights. As shown in Figure 2, in an embodiment, the method includes steps 201 to 203, wherein:

Step 201: Based on several given samples, according to the set binning method, the predicted values of the online prediction model and the predicted values of the offline prediction model are respectively binned, wherein each sample in the several samples includes: A predicted value, a second predicted value and a label of the sample, the first predicted value is predicted by an online prediction model, and the second predicted value is predicted by an offline prediction model.

The samples mentioned in step 201 may be the same as the samples used to train the above-mentioned offline prediction model and/or online prediction model, of course, they may also be different samples, which is not limited.

In an embodiment, the set binning method may be an entropy-based binning method. The entropy-based binning method considers the value of the dependent variable when binning, so that the minimum entropy can be achieved after binning. The benefit of entropy-based binning is that it can show better discrimination in high-scoring regions. Certainly, the set binning method may also be a Gini-based binning method, or an equal frequency binning method, and the like.

Step 202: According to the result of binning, transform the first predicted value of each sample into a first interval feature corresponding to the interval in which the first predicted value is located, and convert the second predicted value of each sample into a feature corresponding to the second interval. The second interval feature corresponding to the interval where the predicted value is located.

In one example, assuming that the first predicted value and the second predicted value are both between 0 and 1, after binning the predicted value of the online prediction model, the obtained segmentation points include: 0, 0.1, 0.13 . The output values of the online prediction model and the offline prediction model are divided into 7 intervals respectively after binning.

In an embodiment, the feature conversion in step 202 can be realized by using the one-hot rule. Assuming that the first predicted value of a sample is 0.17 and the second predicted value is 0.12, since 0.17 is in the fourth interval (0.15, 0.2) and 0.12 is in the sixth interval (0.11, 0.13), one-hot The rule can convert the first predicted value: 0.17 to the first interval feature: on-bin-0001000 ("on-bin" is the identifier of the online prediction model), and convert the second predicted value: 0.12 to the second interval feature: off -bin-0000010 ("off-bin" is the identifier of the offline prediction model). According to the same method, feature transformation can be performed on the first predicted value and the second predicted value in other samples one by one.

Step 203: Use the first interval feature, the second interval feature and the label of the sample corresponding to each sample to form the converted sample data, and use the converted sample data to train the model. The trained model is used It is used to fuse the prediction value of the online prediction model and the prediction value of the offline prediction model to obtain the final prediction value.

Wherein, the converted sample data may also include other data besides the first interval feature, the second interval feature and the label of the sample. That is, the "composition" is not closed.

In the above example, before feature conversion, a piece of sample data is, for example:

{0.17, 0.12, "Medium Risk User"};

After feature conversion, the new piece of sample data obtained is, for example:

{0001000, 0000010, "Medium Risk User"}

The model to be trained herein can be a linear model or a nonlinear model. In an embodiment using a linear model, the parameters to be trained of the model can include weights corresponding to each interval obtained by binning, and the weights can be used The final prediction value is obtained by merging the prediction value of the line prediction model and the prediction value of the offline prediction model. The model to be trained can be a Logistic Regression (LR) model, wherein a weight can be assigned to each interval obtained by binning, and the weight can be used as a parameter of the LR model for training, and finally each weight value can be obtained . The above weight can be a score for the corresponding interval. The score is not only between different model features (online and offline models), but also a global importance trade-off and learning between each score interval.

Following the example mentioned above, the following weights can finally be obtained:

The weight of interval (0,0.1) on-bin-1=1.054,

...

The weight of interval (0.5,1) on-bin-7=4.439;

The weight of interval (0,0.03) off-bin-1=0.604,

...

The weight off-bin-7=3.237 for the interval (0.13,1).

Next, continue to describe the above steps 103 to 104 in conjunction with the above specific examples. Assuming that for a certain target user, the first predicted value obtained through the online prediction model is 0.66, and the second predicted value obtained through the offline prediction model is 0.25, then in combination with the above example, first in step 103, the first predicted value is determined The first interval of the value 0.4 is: (0.5, 1), and the second interval of the second predicted value 0.25 is: (0.13, 1). Then in step 1041, based on the predetermined weights corresponding to each interval obtained by binning, the first weight corresponding to the first interval: (0.5, 1) can be obtained: 4.439, and the second interval : (0.13, 1) corresponds to the second weight: 3.237.

Finally, in step 1042, the final fused prediction value may be determined according to the first weight and the second weight, and in an optional embodiment, the first weight and the second weight may be summed, And the summation result is used as the fusion prediction value, that is, the fusion prediction value=4.439+3.237=7.676. Of course, the specific way of fusion is not limited to summation, such as averaging. Finally, how to use the fused prediction value can be decided according to the specific business.

The effects produced by the technical solutions provided in the embodiments of this specification include:

The predicted value of the line prediction model and the predicted value of the offline prediction model are fused through the weight obtained by machine learning, and finally the score obtained by fusion is used to predict the user's label, thereby improving the user's While predicting the accuracy of tags, it also meets the business requirements for low latency. In addition, using entropy-based binning and logistic regression models, the online model scores and offline model scores are effectively integrated, so that the comparability between online and offline scores can be adaptively adjusted during the machine learning process.

Corresponding to the foregoing method embodiments, the embodiments of this specification further provide an apparatus for fusing model prediction values.

Referring to FIG. 3, in an embodiment, in the training phase of fusion weights, a device 300 for determining fusion weights may include:

The binning unit 301 is configured to bin the predicted values of the online prediction model and the predicted values of the offline prediction model according to a set binning method based on several given samples, wherein, among the several samples Each sample includes: a first predicted value, a second predicted value and a label of the sample, the first predicted value is predicted by an online prediction model, and the second predicted value is predicted by an offline prediction model;

The feature conversion unit 302 is configured to: convert the first predicted value of each sample into a first interval feature corresponding to the interval in which the first predicted value is located according to the binning result, and convert the second predicted value of each sample into Converted into the second interval feature corresponding to the interval where the second predicted value is located;

The training unit 303 is configured to: use the first interval feature, the second interval feature and the label of the sample corresponding to each sample to form converted sample data, and use the converted sample data to train the model, the The trained model is used to fuse the prediction value of the online prediction model and the prediction value of the offline prediction model to obtain the final prediction value.

Referring to FIG. 4, in an embodiment, in the score fusion stage, an apparatus 400 for fusing model prediction values may include:

The online score prediction unit 401 is configured to: acquire the business data generated by the target user within the first time period before the trigger moment, determine the input features according to the business data and input them into the online prediction model, and output the first predicted value, The online prediction model is used to predict the label of the user;

The offline score obtaining unit 402 is configured to: obtain a second predicted value corresponding to the target user obtained by using an offline prediction model, wherein the input characteristics of the offline prediction model are based on the target user's past first determined by the business characteristics generated within two time periods, and the offline prediction model is used to predict the label of the user;

The interval determination unit 403 is configured to: respectively determine the first interval where the first prediction value is located and the second interval where the first prediction value is located according to the results of binning the prediction values of the online prediction model and the prediction values of the offline prediction model in advance. The second interval where the predicted value is located;

The weight determining unit 404 is configured to: according to the first interval and the second interval, use a pre-trained model to fuse the first predicted value and the second predicted value to obtain a final fusion A predicted value, the fused predicted value is used to determine the label of the target user.

In an optional embodiment, the score fusion unit 404 may include:

The weight determining subunit obtains the first weight corresponding to the first interval and the second weight corresponding to the second interval based on the predetermined weight corresponding to each interval obtained by binning;

The fusion subunit is configured to determine a fusion prediction value by using the first weight and the second weight, and the fusion prediction value is used to determine the label of the target user.

In an embodiment, the fusion subunit may be configured as:

Summing the first weight and the second weight, and using the sum result as a fusion prediction value.

For the implementation process of the functions and effects of each module in the above-mentioned device, please refer to the implementation process of the corresponding steps in the above-mentioned method for details, and details will not be repeated here.

The embodiment of this specification also provides a computer device (such as a server), which at least includes a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein the aforementioned method is implemented when the processor executes the program .

FIG. 5 shows a schematic diagram of a more specific hardware structure of a computing device provided by the embodiment of this specification. The device may include: a processor 1010 , a memory 1020 , an input/output interface 1030 , a communication interface 1040 and a bus 1050 . The processor 1010 , the memory 1020 , the input/output interface 1030 and the communication interface 1040 are connected to each other within the device through the bus 1050 .

The processor 1010 may be implemented by a general-purpose CPU (Central Processing Unit, central processing unit), a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute related programs to realize the technical solutions provided by the embodiments of this specification.

The memory 1020 may be implemented in the form of ROM (Read Only Memory, read only memory), RAM (Random Access Memory, random access memory), static storage device, dynamic storage device, and the like. The memory 1020 can store operating systems and other application programs. When implementing the technical solutions provided by the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1020 and invoked by the processor 1010 for execution.

The input/output interface 1030 is used to connect the input/output module to realize information input and output. The input/output/module can be configured in the device as a component (not shown in the figure), or can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, mouse, touch screen, microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, and the like.

The communication interface 1040 is used to connect a communication module (not shown in the figure), so as to realize the communication interaction between the device and other devices. The communication module can realize communication through wired means (such as USB, network cable, etc.), and can also realize communication through wireless means (such as mobile network, WIFI, Bluetooth, etc.).

Bus 1050 includes a path that carries information between the various components of the device (eg, processor 1010, memory 1020, input/output interface 1030, and communication interface 1040).

It should be noted that although the above device only shows the processor 1010, the memory 1020, the input/output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components. In addition, those skilled in the art can understand that the above-mentioned device may only include components necessary to implement the solutions of the embodiments of this specification, and does not necessarily include all the components shown in the figure.

It can be known from the above description of the implementation manners that those skilled in the art can clearly understand that the embodiments of this specification can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the essence of the technical solutions of the embodiments of this specification or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in storage media, such as ROM/RAM, A magnetic disk, an optical disk, etc., include several instructions to enable a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this specification.

The systems, devices, modules, or units described in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementing device is a computer, which may take the form of a personal computer, laptop computer, cellular phone, camera phone, smart phone, personal digital assistant, media player, navigation device, e-mail device, game control device, etc. desktops, tablets, wearables, or any combination of these.

Each embodiment in this specification is described in a progressive 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 device embodiments described above are only illustrative, and the modules described as separate components may or may not be physically separated, and the functions of each module may be integrated in the same or multiple software and/or hardware implementations. Part or all of the modules can also be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without creative effort.

The above is only the specific implementation of the embodiment of this specification. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the embodiment of this specification, some improvements and modifications can also be made. These Improvements and modifications should also be regarded as the scope of protection of the embodiments of this specification.

Claims (14)

1. a kind of method merged to model predication value, including：

Based on given several samples, come the predicted value to on-line prediction model respectively and offline prediction mould according to setting branch mailbox method The predicted value of type carries out branch mailbox, wherein, each sample in several samples includes：First predicted value, the second predicted value with And the label of sample, first predicted value are obtained by on-line prediction model prediction, the second predicted value is pre- by offline prediction model It measures；

According to branch mailbox as a result, the first predicted value of each sample is converted into corresponding with the section residing for first predicted value It is special to be converted into second interval corresponding with the section residing for second predicted value by one section feature for second predicted value of each sample Sign；

After conversion being formed with the label of the corresponding first interval feature of each sample, the second interval feature and sample Sample data, and using the sample data after conversion come training pattern, the model which completes is used for on-line prediction mould The predicted value of type and the predicted value of offline prediction model are merged to obtain final predicted value.

2. according to the method described in claim 1, the setting branch mailbox method includes：Branch mailbox method based on entropy or based on Geordie Branch mailbox method waits frequency divisions case method.

3. according to the method described in claim 1, the model treats training parameter including corresponding with each section that branch mailbox obtains Weight, the weight is used to that the predicted value of line prediction model and the predicted value of offline prediction model to be merged to obtain final Predicted value.

4. a kind of method merged to model predication value, including：

The business datum that is generated in first time period of target user is obtained, input feature vector and defeated is determined according to the business datum Enter to on-line prediction model, export the first predicted value；

The second predicted value corresponding with the target user obtained using offline prediction model is obtained, wherein, it is described offline pre- The input feature vector for surveying model is the service feature generated according to the target user in second time period come definite；

It obtains to the second predicted value progress branch mailbox of the first predicted value and offline prediction model of on-line prediction model as a result, dividing First interval that Que Ding be residing for first predicted value and the second interval residing for second predicted value；

According to the first interval and the second interval, the model that is obtained using advance training come to first predicted value and Second predicted value is merged, and obtains final fusion forecasting value, and the fusion forecasting value is used for determining that the target is used The label at family.

5. according to the method described in claim 3, the model obtained using advance training come to first predicted value and Second predicted value is merged to obtain final fusion forecasting value, including：

Based on predetermined weight corresponding with branch mailbox obtains each section, the first power corresponding with the first interval is obtained Weight and the second weight corresponding with the second interval, the model treat that training parameter includes each section pair obtained with branch mailbox The weight answered；

Fusion forecasting value is determined using first weight and second weight.

6. according to the method described in claim 5, described determine that fusion is pre- using first weight and second weight Measured value, including：

First weight and second weight are summed, and using summed result as fusion forecasting value.

7. a kind of device merged to model predication value, including：

Branch mailbox unit, based on given several samples, according to setting branch mailbox method come the predicted value to on-line prediction model respectively and The predicted value of offline prediction model carries out branch mailbox, wherein, each sample in several samples includes：First predicted value, The label of two predicted values and sample, first predicted value are obtained by on-line prediction model prediction, and the second predicted value is by offline Prediction model is predicted to obtain；

Feature Conversion unit, according to residing for branch mailbox as a result, being converted into the first predicted value of each sample with first predicted value The corresponding first interval feature in section, the second predicted value of each sample is converted into and the section pair residing for second predicted value The second interval feature answered；

Training unit, with the label of the corresponding first interval feature of each sample, the second interval feature and sample Form the sample data after conversion, and using the sample data after conversion come training pattern, the model which completes for pair The predicted value of on-line prediction model and the predicted value of offline prediction model are merged to obtain final predicted value.

8. device according to claim 7, the setting branch mailbox method includes：Branch mailbox method based on entropy or based on Geordie Branch mailbox method waits frequency divisions case method.

9. device according to claim 7, the model treats training parameter including corresponding with each section that branch mailbox obtains Weight, the weight is used to that the predicted value of line prediction model and the predicted value of offline prediction model to be merged to obtain final Predicted value.

10. a kind of device merged to model predication value, including：

Online score value predicting unit obtains the business datum that target user generates in the first time period before triggering moment, root Input feature vector is determined according to the business datum and is input to on-line prediction model, exports the first predicted value, the on-line prediction mould Type is used to predict the label of user；

Offline score value obtaining unit obtains the second prediction corresponding with the target user obtained using offline prediction model Value, wherein, the input feature vector of the offline prediction model is to be generated according to the target user in past second time period Service feature come definite, the offline prediction model is used to predict the label of user；

Interval determination unit carries out branch mailbox according to the predicted value in advance to on-line prediction model and the predicted value of offline prediction model As a result, respectively determine first predicted value residing for first interval and the second interval residing for second predicted value；

Score value integrated unit, according to the first interval and the second interval, the model obtained using advance training is come to institute It states the first predicted value and second predicted value is merged, obtain final fusion forecasting value, the fusion forecasting value is used for Determine the label of the target user.

11. device according to claim 10, the score value integrated unit includes：

Weight determination subelement based on predetermined weight corresponding with branch mailbox obtains each section, obtains and described first Corresponding first weight in section and the second weight corresponding with the second interval；

Subelement is merged, fusion forecasting value, the fusion forecasting value are determined using first weight and second weight For determining the label of the target user.

12. according to the devices described in claim 11, the fusion subelement is configured as：

First weight and second weight are summed, and using summed result as fusion forecasting value.

13. a kind of computer equipment, including：

Processor；

For storing the memory of processor-executable instruction；

The processor is configured as：

Based on given several samples, come the predicted value to on-line prediction model respectively and offline prediction mould according to setting branch mailbox method The predicted value of type carries out branch mailbox, wherein, each sample in several samples includes：First predicted value, the second predicted value with And the label of sample, first predicted value are obtained by on-line prediction model prediction, the second predicted value is pre- by offline prediction model It measures；

According to branch mailbox as a result, the first predicted value of each sample is converted into corresponding with the section residing for first predicted value It is special to be converted into second interval corresponding with the section residing for second predicted value by one section feature for second predicted value of each sample Sign；

After conversion being formed with the label of the corresponding first interval feature of each sample, the second interval feature and sample Sample data, and using the sample data after conversion come training pattern, the model which completes is used for on-line prediction mould The predicted value of type and the predicted value of offline prediction model are merged to obtain final predicted value.

14. a kind of computer equipment, including：

Processor；

For storing the memory of processor-executable instruction；

The processor is configured as：

The business datum that is generated in first time period of target user is obtained, input feature vector and defeated is determined according to the business datum Enter to on-line prediction model, export the first predicted value；

The second predicted value corresponding with the target user obtained using offline prediction model is obtained, wherein, it is described offline pre- The input feature vector for surveying model is the service feature generated according to the target user in second time period come definite；

It obtains to the second predicted value progress branch mailbox of the first predicted value and offline prediction model of on-line prediction model as a result, dividing First interval that Que Ding be residing for first predicted value and the second interval residing for second predicted value；

According to the first interval and the second interval, the model that is obtained using advance training come to first predicted value and Second predicted value is merged, and obtains final fusion forecasting value, and the fusion forecasting value is used for determining that the target is used The label at family.