A data processing method, apparatus and device

Abstract

The embodiment of the specification discloses a data processing method, device and equipment, the method comprises: obtaining service data of a preset service; performing feature extraction on the service data to obtain features corresponding to the service data, wherein the features corresponding to the service data include features with abnormal feature values; based on the feature value abnormality in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, generating model feature values of a target model applied to the preset service for each feature item in the features corresponding to the service data; inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

Description

TECHNICAL FIELD

[0001] The present document relates to the technical field of computer, and particularly relates to a data processing method, device and equipment. BACKGROUND

[0002] With the continuous development of online business, more merchants will access at a certain time node in the future, but there are certain differences in the self-capability of different merchants to collect transaction information, which leads to different merchants that may have different degrees of missing phenomenon in different fields. Field missing will lead to the feature value used by the specified model deployed in the online business for model prediction being an abnormal value, which will also lead to different degrees of performance reduction phenomenon when using the same global model for prediction for different merchants.

[0003] At present, the research on performance reduction phenomenon caused by field missing is limited to the attack stage, that is, by artificially making some fields missing, the model is developed in the direction of expected performance reduction. Moreover, the feature value abnormality caused by field missing can be simulated in the model training process without modifying any model structure. The above-mentioned method cannot make more accurate prediction for different missing conditions, which also leads to the reduction of the effect of other models that do not consider the field missing condition in model training. Therefore, it is necessary to provide a technical solution that can enable the model to flexibly select the feature that should be paid more attention to for model prediction according to the field missing condition, thereby reducing the influence of field missing on the effect of the model. SUMMARY

[0004] The purpose of the embodiments of the present specification is to provide a technical solution that can enable the model to flexibly select the feature that should be paid more attention to for model prediction according to the field missing condition, thereby reducing the influence of field missing on the effect of the model.

[0005] In order to achieve the above technical solution, the embodiments of the present specification are implemented as follows:

[0006] The data processing method provided by the embodiments of the present specification comprises: obtaining business data of a preset business. Feature extraction is performed on the business data to obtain features corresponding to the business data, wherein the features corresponding to the business data include features with abnormal feature values. Based on the feature value abnormality in the features corresponding to the business data and the features with no abnormal feature values in the features corresponding to the business data, a model feature value of a target model applied in the preset business is generated for each feature item in the features corresponding to the business data. Each feature item in the features corresponding to the business data and the corresponding model feature value are input into the target model to obtain a prediction result corresponding to the business data.

[0007] The embodiment of the present specification provides a data processing device, the device comprises: a data acquisition module, acquiring service data of a preset service. A feature extraction module extracts features of the service data, obtains features corresponding to the service data, and the features corresponding to the service data include features with abnormal feature values. A feature processing module generates model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data based on the abnormal feature value situation in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data. A prediction module inputs each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

[0008] The embodiment of the present specification provides a data processing device, the device comprises: a data acquisition module, acquiring service data of a preset service. A feature extraction module extracts features of the service data, obtains features corresponding to the service data, and the features corresponding to the service data include features with abnormal feature values. A feature processing module generates model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data based on the abnormal feature value situation in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data. A prediction module inputs each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

[0009] The embodiment of the present specification provides a data processing device, the device comprises: a data acquisition module, acquiring service data of a preset service. A feature extraction module extracts features of the service data, obtains features corresponding to the service data, and the features corresponding to the service data include features with abnormal feature values. A feature processing module generates model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data based on the abnormal feature value situation in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data. A prediction module inputs each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to make the technical solutions in the embodiments of the present specification or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present specification, and those skilled in the art can obtain other accompanying drawings according to these accompanying drawings without any creative effort.

[0011] Figure 1 An embodiment of a data processing method of the present specification;

[0012] Figure 2 Another embodiment of a data processing method of the present specification;

[0013] Figure 3 A schematic diagram of a data processing process of the present specification;

[0014] Figure 4 Still another embodiment of a data processing method of the present specification;

[0015] Figure 5 An embodiment of a data processing device of the present specification;

[0016] Figure 6 An embodiment of a data processing equipment of the present specification. DETAILED DESCRIPTION

[0017] The embodiments of the present specification provide a data processing method, device and equipment.

[0018] In order to make the technical solutions in the embodiments of the present specification, the technical solutions in the embodiments of the present specification will be described clearly and completely below with reference to the accompanying drawings of the embodiments of the present specification. Obviously, the described embodiments are only some embodiments of the present specification, not all embodiments. Based on the embodiments in the present specification, all other embodiments obtained by those skilled in the art without any creative effort shall fall within the scope of protection of the present specification.

[0019] Embodiment one

[0020] As Figure 1As shown, the embodiment of the present specification provides a data processing method, the execution subject of the method can be a terminal device or a server, etc., wherein the terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, etc., can also be a computer device such as a notebook computer or a desktop computer, or can also be an IoT device (specifically such as a smart watch, a vehicle-mounted device, etc.), etc., wherein the server can be an independent server, can also be a server cluster composed of multiple servers, etc., the server can be a background server of a financial service or a network shopping service, etc., can also be a background server of an application program, etc. The method can specifically include the following steps:

[0021] In step S102, the service data of the preset service is acquired.

[0022] Among them, the preset service can be any service, for example, the preset service can be a payment service, a transfer service or an instant messaging service, etc., which can be set according to actual conditions, and the present specification does not limit it. The service data can be data related to the preset service, which can be determined according to the preset service, for example, the preset service is a payment service, and the service data can include payment time, place, payment amount, account information of the payer, account information of the receiver, etc., which can be set according to actual conditions, and the present specification does not limit it.

[0023] In implementation, as the online business is continuously developed, more merchants will access at a certain time node in the future, but the self-ability of different merchants to collect transaction information is different, which leads to the fact that different merchants may have different degrees of missing phenomenon in the transmission of different fields (for example, the A merchant may miss 50% of the buyer's device identifier, and the B merchant may miss 70% of the buyer's client identifier), and the field missing will cause the feature value used by the specified model deployed in the online business to be an abnormal value when the model prediction is performed, which will also cause the performance of using the same global model for prediction to be reduced to different degrees for different merchants. However, it is too expensive to maintain a model for each merchant, therefore, it is necessary to solve the performance reduction of the general model caused by the feature value abnormality due to the field missing. The current research on the performance reduction caused by the field missing is limited to the attack stage, that is, by artificially making some fields missing, the model is caused to develop in the direction of the expected performance reduction, and the feature value abnormality caused by the field missing can be simulated in the model training process without modifying any model structure. The above-mentioned manner cannot make more accurate prediction for different missing conditions, and also causes the model effect without considering the field missing condition in the model training to be reduced, therefore, a technical solution is needed, which can enable the model to flexibly select the features that should be paid more attention to according to the field missing condition for model prediction, thereby reducing the influence of the field missing on the model effect. The embodiments of the present specification provide an implementable technical solution, which can be specifically referred to the following content.

[0024] When the specified model is deployed in a certain business (that is, a preset business), if there is a situation that the feature value corresponding to a certain feature item input into the model is missing in the preset business, when a user triggers the preset business to execute, the business data generated in the process of the user executing the preset business can be acquired. Alternatively, a certain amount of business data can be acquired from the business data (or from a specified historical database) recorded in advance in the preset business as training samples in the process of training the model, so as to subsequently train the model.

[0025] In step S104, the business data is subjected to feature extraction to obtain the features corresponding to the business data, and the features corresponding to the business data include the features with abnormal feature values.

[0026] In the embodiment, the features can be divided into feature items and feature values for convenience of description. The feature item can be an identifier of a feature, for example, the feature item can be a feature name or a feature code, and the like. For example, the feature item is a transaction amount, and the corresponding feature value is 100. For another example, the feature item is an account name, and the corresponding feature value is Ac55hd, and the like. The feature values can be set according to actual conditions, and the embodiments of the present specification do not limit the feature values. The feature value anomaly can include a feature value missing, a feature value error, and the like. The feature values can be set according to actual conditions, and the embodiments of the present specification do not limit the feature values.

[0027] In the implementation, after obtaining the service data of the preset service by the above method, a pre-set feature extraction algorithm can be obtained, and the feature extraction algorithm can be used to extract features of the service data, to obtain the features corresponding to the service data. Since the feature values corresponding to one or more different feature items in the service data can be abnormal, for example, the feature value corresponding to a feature item is obviously incorrect (for example, the feature item is a transaction amount, and the corresponding feature value is information such as account A or Bdcge (i.e., English characters) that is irrelevant to the amount), and / or the feature value corresponding to a feature item is missing, for example, the feature item is an identifier of a terminal device used by a payer, and the corresponding feature value is empty (i.e., missing), for another example, the feature item is an IP address of a terminal device used by a payer, and the corresponding feature value is empty (i.e., missing), and the like. The features can be set according to actual conditions. Therefore, the features obtained by the above method can include features with abnormal feature values. If the service data actually includes abnormal data, the features corresponding to the service data can include features with abnormal feature values. The features with abnormal feature values can be features corresponding to the abnormal data.

[0028] It should be noted that the features with abnormal feature values can be feature values that are lost or do not exist in actual applications. The features with abnormal feature values can also be feature values that are set to abnormal values or are emptied by a technician for the purpose of verifying the effect of a model, and the like. The features with abnormal feature values can be set according to actual conditions, and the embodiments of the present specification do not limit the features with abnormal feature values.

[0029] In step S106, based on the feature value anomaly in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, a model feature value of a target model applied to a preset service is generated for each feature item in the features corresponding to the service data.

[0030] The feature value abnormality can include various types, such as a type of feature value abnormality (for example, a feature value abnormality belonging to a feature value missing type or a feature value abnormality belonging to a feature value error type, etc.), an influence degree of a feature value abnormality on other features, a correlation between a feature value abnormality and other features, whether a feature value abnormality belongs to a feature set in which a feature value abnormality frequently occurs (that is, a feature set in which a probability of a feature value abnormality is greater than a preset probability threshold), etc. The target model can be any model, and can be set according to a preset business. For example, when the preset business is a payment business, the target model can be a risk prevention and control model, which can be used to detect whether there is a fraud risk in the process of performing the payment business. Alternatively, when the preset business is an online shopping business, the target model can be an information recommendation model or an information retrieval model, which can be used to recommend or retrieve specified information to a user of the online shopping business. The target model can be constructed by various algorithms, such as a classification algorithm, a neural network model, a genetic algorithm, an ant colony algorithm, etc. The model feature value can be a feature value that can be directly input into the target model, and the target model can accurately identify the content of the model feature value and output a corresponding result. That is, the model feature value can be a feature that does not need to be processed and can be directly applied to the target model.

[0031] In implementation, considering that the feature value abnormality may occur on each feature, in order to improve the effect of the target model, which feature the target model should focus on can be determined according to the input data of the target model, the model feature value of the target model can be re-determined through the above manner, specifically, the features corresponding to the above business data can be analyzed to determine the features containing feature values and the features not containing feature values, and the features containing feature values can be further analyzed to determine the features containing feature value errors, etc., through the above manner, the features not containing feature values and the features containing feature value errors, etc. can be obtained, so that the features containing feature value abnormality in the features corresponding to the above business data can be obtained. The feature value abnormality of the features containing feature value abnormality can be further analyzed to obtain the feature value abnormality of the features containing feature value abnormality. The corresponding algorithm can be pre-set according to the actual situation, for example, the algorithm corresponding to the attention mechanism, etc. The attention mechanism can include multiple, for example, temporal attention mechanism or spatial attention mechanism, or it can also be soft attention or hard attention, etc. The specific algorithm can be set according to the actual situation. The above algorithm can be used, and the features containing feature value abnormality in the features corresponding to the above business data and the features not containing feature value abnormality in the features corresponding to the above business data can be combined to determine the features that the target model should focus on. In this way, the features that the target model should focus on can be determined through the global data situation of the input data, and then the model feature value of the target model applied in the preset business can be generated for each feature item in the features corresponding to the business data based on the determined features that the target model should focus on. The specific processing process of generating the model feature value of the target model applied in the preset business for each feature item in the features corresponding to the business data through the algorithm corresponding to each attention mechanism can be executed based on the processing mode of the corresponding attention mechanism, which will not be described here.

[0032] In step S108, each feature item in the features corresponding to the above business data and the corresponding model feature value are input into the target model to obtain the prediction result corresponding to the above business data.

[0033] It should be noted that in actual application, the model feature value corresponding to each feature item in the features corresponding to the above business data can also be input into the target model to obtain the prediction result corresponding to the above business data, without the need to input each feature item in the features corresponding to the above business data into the target model for processing. The specific processing mode can be selected according to the actual situation, and the embodiments of the present specification are not limited in this regard.

[0034] The embodiment of the present specification provides a data processing method, by acquiring service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, then, based on the features with abnormal feature values in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data, the model can flexibly select the features that should be paid more attention to according to the field missing condition for model prediction, thereby reducing the influence of field missing on the model effect, compared with the model without any processing, the present scheme can improve the prediction effect of the model on the field missing condition, compared with the way of not modifying the model and directly considering the field missing condition in model training, the present scheme can ensure that the effect of other conditions is not affected, and can greatly reduce the cost of model training and the size of the model, facilitating the deployment of the model and reducing the prediction time cost of the model.

[0035] Embodiment two

[0036] As Figure 2 shown, the embodiment of the present specification provides a data processing method, the execution subject of the method can be a terminal device or a server, etc., wherein the terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, etc., can also be a computer device such as a notebook computer or a desktop computer, or can also be an IoT device (such as a smart watch, a vehicle-mounted device, etc.), etc., wherein the server can be an independent server, can also be a server cluster composed of multiple servers, etc., the server can be a background server of a financial service or an online shopping service, etc., can also be a background server of an application program, etc. The method can specifically include the following steps:

[0037] In step S202, the service data of a preset service is acquired.

[0038] In step S204, feature extraction is performed on the above service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values.

[0039] The feature value abnormality can include feature value missing.

[0040] In step S206, the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the service data based on the feature value abnormality in the feature corresponding to the service data and the feature with no feature value abnormality in the feature corresponding to the service data, and the model feature value is applied to a target model in a preset service.

[0041] The self-attention algorithm is an algorithm that only needs to determine which features should be focused on according to the input. The calculation process of the self-attention algorithm can include: converting the input data into an embedding vector; creating a Query vector, a Key vector, and a Value vector for each embedding vector, wherein the created vectors are obtained by multiplying the embedding vector of the input data and three conversion matrices (W Q, W K, and W V ), and the three matrices are learned in the training process; calculating a corresponding score score for each embedding vector, score = Query vector dot product Key vector; dividing the score by the square root of the dimension of the Key vector, so that the obtained gradient is more stable, and then the score is normalized by a softmax function to make the sum equal to 1; multiplying the score obtained by the softmax function with the corresponding value vector, so as to retain the value of the focused feature and weaken the value of the irrelevant feature; and accumulating all the weighted value vectors, so as to obtain the output result of the self-attention algorithm at a certain position.

[0042] In implementation, the model feature value corresponding to each feature item in the feature corresponding to the service data can be calculated by the calculation process of the self-attention algorithm provided above in combination with the feature value abnormality in the feature corresponding to the service data and the feature with no feature value abnormality in the feature corresponding to the service data, and details are not repeated here.

[0043] The processing of step S206 can be various, and two optional processing modes are provided below, which can include the processing of mode one and mode two.

[0044] Mode one: can include the processing of step A4 in step A2.

[0045] In step A2, the target feature with feature value abnormality in the feature corresponding to the service data is obtained, and the self-attention algorithm set in advance for the target feature is obtained according to the target feature.

[0046] In step A4, using the acquired self-attention algorithm, based on the abnormal feature values ​​in the features corresponding to the above business data and the features with no abnormal feature values ​​in the features corresponding to the business data, model feature values ​​are generated for each feature item in the features corresponding to the business data, which are then applied to the target model in the preset business.

[0047] In implementation, such as Figure 3 As shown, analysis can reveal the features among the above business data whose feature values ​​are normal and those whose feature values ​​are abnormal. Figure 3 Features with normal feature values ​​are represented by "1", and features with abnormal feature values ​​are represented by "0". Specifically, the second and last features from the left in the feature sequence corresponding to the aforementioned business data are considered abnormal features, while the remaining features are considered normal. Then, using the acquired self-attention algorithm and the pre-set self-attention algorithm corresponding to features with normal feature values, combined with the abnormal and normal feature values ​​in the features corresponding to the aforementioned business data, the calculation process of the self-attention algorithm is applied to generate model feature values ​​for each feature item in the features corresponding to the business data, applicable to the target model in the preset business.

[0048] Method 2: It may include the following steps B2 and B4.

[0049] In step B2, target features with abnormal feature values ​​are obtained from the features corresponding to the above business data, and the target features are classified to obtain the category corresponding to each target feature, and the self-attention algorithm corresponding to each category is obtained.

[0050] The categories can include a variety of categories, such as user personal information, terminal device information, amount, time, etc. A corresponding self-attention algorithm can be set in advance for each category. Different categories may have different self-attention algorithms, which can be set according to the actual situation. This specification does not limit this in the embodiments.

[0051] In step B4, the self-attention algorithm corresponding to each category is used to generate model feature values ​​for each feature item in the features corresponding to the business data, based on the abnormal feature values ​​in the features corresponding to the above business data, the category corresponding to each target feature, and the features with no abnormal feature values ​​in the features corresponding to the business data. These model feature values ​​are then applied to the target model in the preset business.

[0052] In implementation, such as Figure 3As shown, the self-attention algorithm obtained and the self-attention algorithm corresponding to the feature value without exception in the preset feature can be combined with the feature value exception in the feature corresponding to the business data, the category corresponding to each target feature, and the feature with no feature value exception in the feature corresponding to the business data, and the calculation process of the self-attention algorithm is used for calculation, so as to generate the model feature value of the target model applied in the preset business for each feature item in the feature corresponding to the business data.

[0053] In step S208, each feature item in the feature corresponding to the business data and the corresponding model feature value are input into the target model to obtain the prediction result corresponding to the business data.

[0054] In step S210, according to the prediction result corresponding to the business data, it is judged whether the prediction result can make the target model converge.

[0055] In step S212, if no, the business data of the preset business continues to train the target model until the target model converges, the trained target model is obtained, and the trained target model is deployed in the preset business.

[0056] It should be noted that the processing of the missing feature value occurs in the training process of the target model, that is, in the process of training the target model, the missing feature value can be processed in the above manner, so as to obtain a target model with better model effect.

[0057] The embodiment of the present specification provides a data processing method, by obtaining service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, and then, based on the feature value abnormality in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data, the model can flexibly select the features that should be paid more attention to according to the field missing condition for model prediction, thereby reducing the influence of field missing on the model effect, compared with the model without any processing, the present scheme can improve the prediction effect of the model on the field missing condition, compared with the way of not modifying the model, but directly considering the field missing condition in the model training, the present scheme can ensure that the effect of other conditions is not affected, and can greatly reduce the cost of model training and the size of the model, facilitating the deployment of the model and reducing the prediction time cost of the model.

[0058] In addition, an improved scheme based on the self-attention algorithm is proposed, which can automatically generate different final model features according to the field missing condition of the model to improve the model performance under different field missing conditions.

[0059] Embodiment three

[0060] As shown in Figure 4 The embodiment of the present specification provides a data processing method, the execution subject of the method can be a terminal device or a server, etc., wherein the terminal device can be a mobile terminal device such as a mobile phone, a tablet computer, etc., and can also be a computer device such as a notebook computer or a desktop computer, or can also be an IoT device (such as a smart watch, a vehicle-mounted device, etc.), etc., wherein the server can be an independent server, and can also be a server cluster composed of multiple servers, etc., the server can be a background server of a financial service or an online shopping service, etc., or can also be a background server of an application program, etc. The method can specifically include the following steps:

[0061] In step S402, a deployment instruction of a target model is obtained, the target model being a model obtained by model training based on service data of a preset service.

[0062] In step S404, based on the above deployment instruction, the target model is obtained, and the target model is deployed in the preset service.

[0063] In step S406, service data of the preset service is acquired.

[0064] In step S408, feature extraction is performed on the service data to obtain features corresponding to the service data, wherein the features corresponding to the service data include features with abnormal feature values.

[0065] The abnormal feature values can include missing feature values.

[0066] In step S410, a self-attention algorithm is used to generate model feature values of each feature item in the features corresponding to the service data based on the abnormal feature values in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, and the model feature values are applied to a target model in the preset service.

[0067] The processing in step S410 can be various, and two optional processing modes are provided below, which can include the processing in mode one and mode two.

[0068] Mode one: can include the processing in step C4 in step C2.

[0069] In step C2, a target feature with abnormal feature values in the features corresponding to the service data is acquired, and a self-attention algorithm previously set for the target feature is acquired according to the target feature.

[0070] In step C4, the self-attention algorithm is used to generate model feature values of each feature item in the features corresponding to the service data based on the abnormal feature values in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, and the model feature values are applied to a target model in the preset service.

[0071] Mode two: can include the processing in step D2 and step D4.

[0072] In step D2, a target feature with abnormal feature values in the features corresponding to the service data is acquired, and the target feature is classified to obtain a category corresponding to each target feature, and a self-attention algorithm corresponding to each obtained category is acquired.

[0073] In step D4, the self-attention algorithm corresponding to each category is used to generate model feature values of each feature item in the features corresponding to the service data based on the abnormal feature values in the features corresponding to the service data, the category corresponding to each target feature, and the features with normal feature values in the features corresponding to the service data, and the model feature values are applied to a target model in the preset service.

[0074] In step S412, each feature item in the features corresponding to the service data and the corresponding model feature value are input into the target model to obtain a prediction result corresponding to the service data.

[0075] The specific processing of steps S402-S412 can be referred to the foregoing related content, which will not be described here.

[0076] It should be noted that the processing of the missing feature value occurs in the process of predicting real-time service data after the target model is deployed in the preset service, that is, after the target model is deployed in the preset service, the missing feature value can be processed in the above manner to obtain a better output result of the target model.

[0077] The embodiment of the present specification provides a data processing method, by obtaining service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, then, based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature value into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data, the model can flexibly select the features that should be paid more attention to for model prediction according to the field missing situation, thereby reducing the influence of field missing on the model effect, compared with the model without any processing, the present scheme can improve the prediction effect of the model on the field missing situation, compared with the way of not modifying the model, but directly considering the field missing situation in the model training, the present scheme can ensure not to affect the effect of other situations, and can greatly reduce the cost of model training and the size of the model, facilitate the deployment of the model and reduce the prediction time cost of the model.

[0078] In addition, an improved scheme based on the self-attention algorithm is proposed, which can automatically generate different final model features according to the field missing situation of the model to improve the model efficiency of the model in different field missing situations.

[0079] Embodiment Four

[0080] The above is the data processing method provided by the embodiment of the present specification, based on the same idea, the embodiment of the present specification also provides a data processing device, as shown in Figure 5 .

[0081] The data processing apparatus comprises a data acquisition module 501, a feature extraction module 502, a feature processing module 503, and a prediction module 504, wherein:

[0082] The data acquisition module 501 acquires service data of a preset service.

[0083] The feature extraction module 502 performs feature extraction on the service data to obtain features corresponding to the service data, wherein the features corresponding to the service data include features with abnormal feature values.

[0084] The feature processing module 503 generates model feature values of each feature item in the features corresponding to the service data for application to a target model in the preset service based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data.

[0085] The prediction module 504 inputs each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

[0086] In the embodiments of the present specification, the feature processing module 503 generates model feature values of each feature item in the features corresponding to the service data for application to a target model in the preset service based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data through a self-attention algorithm.

[0087] In the embodiments of the present specification, the feature processing module 503 comprises:

[0088] The first algorithm acquisition unit acquires a target feature with abnormal feature values in the features corresponding to the service data, and acquires a self-attention algorithm previously set for the target feature according to the target feature.

[0089] The first feature processing unit generates model feature values of each feature item in the features corresponding to the service data for application to a target model in the preset service based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data through the acquired self-attention algorithm.

[0090] In the embodiments of the present specification, the feature processing module 503 comprises:

[0091] The second algorithm acquisition unit acquires target features with abnormal feature values in the features corresponding to the service data, classifies the target features, obtains a category corresponding to each target feature, and acquires a self-attention algorithm corresponding to each obtained category.

[0092] The second feature processing unit generates, for each feature item in the features corresponding to the service data, a model feature value of a target model applied in the preset service by using the self-attention algorithm corresponding to each obtained category, based on the abnormal feature value situation in the features corresponding to the service data, the category corresponding to each target feature, and the feature with no abnormal feature value in the features corresponding to the service data.

[0093] In the embodiments of the present specification, the abnormal feature value includes a missing feature value.

[0094] In the embodiments of the present specification, the apparatus further includes:

[0095] The judging module judges whether the prediction result can make the target model converge according to the prediction result corresponding to the service data.

[0096] The training module, if not, continues to perform model training on the target model by using the service data of the preset service until the target model converges, obtains a trained target model, and deploys the trained target model in the preset service.

[0097] In the embodiments of the present specification, the apparatus further includes:

[0098] The instruction acquisition module acquires a deployment instruction of the target model, and the target model is a model obtained by performing model training on the service data of a preset service.

[0099] The model deployment module acquires the target model based on the deployment instruction, and deploys the target model in the preset service.

[0100] The embodiment of the present specification provides a data processing apparatus, by acquiring service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, then, based on the feature value abnormality in the features corresponding to the service data and the features with normal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data, the model can flexibly select the features that should be paid more attention to according to the field missing condition to make prediction of the model, thereby reducing the influence of field missing on the model effect, compared with the model without any processing, the present scheme can improve the prediction effect of the model on the field missing condition, compared with the way of not modifying the model, but directly considering the field missing condition in the model training, the present scheme can ensure that the effect of other conditions is not affected, and can greatly reduce the cost of model training and the size of the model, facilitating the deployment of the model and reducing the prediction time cost of the model.

[0101] In addition, an improved scheme based on the self-attention algorithm is proposed, which can automatically generate different final model features according to the field missing condition of the model to improve the model performance under different field missing conditions.

[0102] Embodiment five

[0103] The above is the data processing apparatus provided by the embodiment of the present specification, based on the same idea, the embodiment of the present specification also provides a data processing device, as shown in Figure 6 .

[0104] The data processing device can be arranged in the terminal device or the server provided in the above embodiments.

[0105] Data processing devices can vary greatly in configuration and performance, and can include one or more processors 601 and memory 602, which can store one or more stored applications or data. The memory 602 can be volatile or non-volatile memory. The applications stored in the memory 602 can include one or more modules (not shown), each of which can include a series of computer-executable instructions for the data processing device. Further, the processor 601 can be configured to communicate with the memory 602 to execute the series of computer-executable instructions in the memory 602 on the data processing device. The data processing device can also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input / output interfaces 605, and one or more keyboards 606.

[0106] In particular embodiments, a data processing device includes memory and one or more programs, wherein one or more programs are stored in the memory and the one or more programs can include one or more modules, and each module can include a series of computer-executable instructions for the data processing device, and the one or more programs configured to be executed by one or more processors include computer-executable instructions for performing:

[0107] Obtaining service data of a preset service;

[0108] Performing feature extraction on the service data to obtain features corresponding to the service data, wherein the features corresponding to the service data include features with abnormal feature values;

[0109] Based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating model feature values of each feature item in the features corresponding to the service data for a target model in the preset service;

[0110] Inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

[0111] In the embodiments of the present specification, the generating of the model feature values of each feature item in the features corresponding to the service data for the target model in the preset service based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data includes:

[0112] The self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data.

[0113] In the embodiments of the present specification, the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, and applied to a target model in the preset business, including:

[0114] The target feature with feature value abnormality in the feature corresponding to the business data is obtained, and a self-attention algorithm previously set for the target feature is obtained according to the target feature.

[0115] The self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, and applied to a target model in the preset business.

[0116] In the embodiments of the present specification, the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, and applied to a target model in the preset business, including:

[0117] The target feature with feature value abnormality in the feature corresponding to the business data is obtained, and each target feature is classified to obtain a corresponding category, and a self-attention algorithm corresponding to each category is obtained.

[0118] The self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, and applied to a target model in the preset business.

[0119] In the embodiments of the present specification, the feature value abnormality includes feature value missing.

[0120] In the embodiments of the present specification, the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, and applied to a target model in the preset business, including:

[0121] determine whether the prediction result can make the target model converge according to a prediction result corresponding to the service data;

[0122] If not, service data of the preset service is acquired to continue model training of the target model until the target model converges, a trained target model is obtained, and the trained target model is deployed in the preset service.

[0123] In the embodiments of the present specification, before the service data of the preset service is acquired, the method further includes:

[0124] acquiring a deployment instruction of the target model, the target model being a model obtained through model training of service data of a preset service;

[0125] based on the deployment instruction, acquiring the target model and deploying the target model in the preset service.

[0126] The embodiments of the present specification provide a data processing device, by acquiring service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, then, based on the features with abnormal feature values in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating, for each feature item in the features corresponding to the service data, a model feature value applied to a target model in a preset service, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature value into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating, for each feature item in the features corresponding to the service data, a model feature value applied to a target model in a preset service, the model can flexibly select features that should be paid more attention to for model prediction according to the field missing situation, thereby reducing the impact of field missing on the model effect, compared with a model without any processing, the present scheme can improve the prediction effect of the model on the field missing situation, compared with a method of not modifying the model and directly considering the field missing situation in model training, the present scheme can ensure that the effect of other situations is not affected, and can greatly reduce the cost of model training and the size of the model, facilitate the deployment of the model and reduce the prediction time cost of the model.

[0127] In addition, an improved scheme based on the self-attention algorithm is proposed, which can automatically generate different final model features according to the field missing situation of the model to improve the model efficiency of the model in different field missing situations.

[0128] Embodiment six

[0129] Further, based on the above Figures 1 to 4The method shown, one or more embodiments of the present specification also provides a storage medium for storing computer executable instruction information, in a specific embodiment, the storage medium can be a U disk, a CD, a hard disk, etc., the computer executable instruction information stored in the storage medium can realize the following process when executed by the processor:

[0130] obtaining service data of a preset service;

[0131] performing feature extraction on the service data to obtain features corresponding to the service data, wherein the features corresponding to the service data include features with abnormal feature values;

[0132] based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data;

[0133] inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data.

[0134] In an embodiment of the present specification, the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data include:

[0135] based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data by using a self-attention algorithm.

[0136] In an embodiment of the present specification, the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data by using a self-attention algorithm include:

[0137] obtaining a target feature with an abnormal feature value in the features corresponding to the service data, and obtaining a self-attention algorithm previously set for the target feature according to the target feature;

[0138] The self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, which is applied to the target model in the preset business, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data.

[0139] In the embodiments of the present specification, the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data, which is applied to the target model in the preset business, based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, including:

[0140] The target feature with feature value abnormality in the feature corresponding to the business data is obtained, and the target feature is classified to obtain a category corresponding to each target feature, and a self-attention algorithm corresponding to each obtained category is obtained.

[0141] The self-attention algorithm corresponding to each category is used to generate a model feature value of each feature item in the feature corresponding to the business data, which is applied to the target model in the preset business, based on the feature value abnormality in the feature corresponding to the business data, the category corresponding to each target feature, and the feature with no feature value abnormality in the feature corresponding to the business data.

[0142] In the embodiments of the present specification, the feature value abnormality includes feature value missing.

[0143] In the embodiments of the present specification, it further includes:

[0144] According to the prediction result corresponding to the business data, it is judged whether the prediction result can make the target model converge;

[0145] If not, the business data of the preset business is obtained to continue model training of the target model until the target model converges, a trained target model is obtained, and the trained target model is deployed in the preset business.

[0146] In the embodiments of the present specification, before obtaining the business data of the preset business, it further includes:

[0147] The deployment instruction of the target model is obtained, and the target model is obtained by model training of the business data of the preset business;

[0148] Based on the deployment instruction, the target model is obtained, and the target model is deployed in the preset business.

[0149] The embodiment of the present specification provides a storage medium, by acquiring service data of a preset service, performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including features with abnormal feature values, then, based on the feature value abnormality in the features corresponding to the service data and the features with no abnormal feature values in the features corresponding to the service data, generating model feature values of a target model applied in the preset service for each feature item in the features corresponding to the service data, finally, inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the target model to obtain a prediction result corresponding to the service data, in this way, by regenerating the model feature values of the target model applied in the preset service for each feature item in the features corresponding to the service data, the model can flexibly select features that should be paid more attention to according to the field missing condition to make prediction of the model, thereby reducing the influence of field missing on the model effect, compared with the model without any processing, the present scheme can improve the prediction effect of the model on the field missing condition, compared with the way of not modifying the model and directly considering the field missing condition in the model training, the present scheme can ensure that the effect of other conditions is not affected, and can greatly reduce the cost of model training and the size of the model, facilitate the deployment of the model and reduce the prediction time cost of the model.

[0150] In addition, an improved scheme based on the self-attention algorithm is proposed, which can automatically generate different final model features according to the field missing condition of the model to improve the model performance in different field missing conditions.

[0151] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be utilized or can be advantageous.

[0152] In the 1990s, it was relatively easy to distinguish whether an improvement in a technology was a hardware improvement (e.g., an improvement in the circuit structure of a diode, transistor, switch, etc.) or a software improvement (an improvement in a method flow). However, as technology has evolved, many improvements in method flows today can be considered as direct improvements in hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structures by programming the improved method flows into hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using hardware entity modules. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device. A digital system is "integrated" on a PLD by the designer programming the PLD, rather than by ordering a chip manufacturer to design and fabricate a custom integrated circuit chip. Moreover, instead of manually fabricating integrated circuit chips, this programming is now mostly implemented using "logic compiler" software, which is similar to software compilers used in program development, and the original code to be compiled is written in a specific programming language, which is called a hardware description language (HDL), and there are many such languages, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., and the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should be aware that, as long as the method flow is logically programmed in the above-mentioned hardware description languages and programmed into an integrated circuit, a hardware circuit implementing the logical method flow can be easily obtained.

[0153] The controller can be implemented in any suitable way, e.g. the controller can take the form of a microprocessor or processor and a computer readable medium storing computer readable program code, e.g. software or firmware, executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller and an embedded microcontroller, examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91 SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to being implemented in pure computer readable program code form, the controller can perfectly well be implemented by means of logic programmed into logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc. to perform the same functions. The controller can thus be considered as a hardware component, and the means comprised therein for performing various functions can be considered as structures within the hardware component. Alternatively, or even, the means for performing various functions can be considered as both a software module implementing a method and a structure within a hardware component.

[0154] The systems, apparatuses, modules or units illustrated by the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0155] For the sake of description, the above apparatuses are described in functional division and are described respectively. Of course, the functions of each unit can be implemented in the same or more software and / or hardware when implementing one or more embodiments of the present specification.

[0156] Those skilled in the art will understand that the embodiments of the present specification can be provided as a method, a system or a computer program product. Therefore, one or more embodiments of the present specification can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0157] The embodiments of the present specification are described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to the embodiments of the present specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable electronic devices to produce a machine, so that the instructions executed by the computer or other programmable electronic devices generate a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks. Figure 1 one or more flows and / or blocks.

[0158] These computer program instructions can also be stored in a computer readable memory capable of directing the computer or other programmable electronic devices to work in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including instruction devices that implement the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks. Figure 1 one or more flows and / or blocks.

[0159] These computer program instructions can also be loaded into a computer or other programmable electronic devices, so that a series of operation steps are performed on the computer or other programmable electronic devices to produce a computer implemented process, so that the instructions executed on the computer or other programmable electronic devices provide steps for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks. Figure 1 one or more flows and / or blocks.

[0160] In a typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces and memories.

[0161] The memory can include non-persistent memory in the computer readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of the computer readable medium.

[0162] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0163] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, such that processes, methods, articles or devices that comprise a list of elements not only include those elements, but also include other elements not expressly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device that includes the element.

[0164] Those skilled in the art will appreciate that embodiments of the present specification can be provided as methods, systems or computer program products. Therefore, one or more embodiments of the present specification can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0165] One or more embodiments of the present specification can be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present specification can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media, including storage devices.

[0166] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0167] The above only describes the embodiments of the specification and is not used to limit the application. The specification can have various changes and modifications for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the specification shall be included in the scope of claims of the specification.

Claims

1. A data processing method, the method comprising: obtaining service data of a preset service, the preset service being a payment service, the service data including payment time, location, payment amount, account information of a payment party, and account information of a receiving party; performing feature extraction on the service data to obtain features corresponding to the service data, the features corresponding to the service data including feature items and feature values corresponding to the feature items, and the features corresponding to the service data including features with abnormal feature values; generating, based on feature value abnormality conditions in the features corresponding to the service data and features without abnormal feature values in the features corresponding to the service data, model feature values of each feature item in the features corresponding to the service data for a risk control model applied in the preset service, the feature value abnormality conditions including what form of feature value abnormality, an influence degree of the feature with abnormal feature values on other features, an association between the feature with abnormal feature values and other features, and features in a feature set with a probability of feature value abnormality greater than a preset probability threshold; inputting each feature item in the features corresponding to the service data and the corresponding model feature values into the risk control model to obtain a prediction result corresponding to the service data.

2. The method of claim 1, wherein the generating, based on the feature value abnormality conditions in the features corresponding to the service data and the features without abnormal feature values in the features corresponding to the service data, model feature values of each feature item in the features corresponding to the service data for a risk control model applied in the preset service comprises: generating, based on the feature value abnormality conditions in the features corresponding to the service data and the features without abnormal feature values in the features corresponding to the service data, model feature values of each feature item in the features corresponding to the service data for a risk control model applied in the preset service by a self-attention algorithm.

3. The method of claim 2, wherein the generating, based on the feature value abnormality conditions in the features corresponding to the service data and the features without abnormal feature values in the features corresponding to the service data, model feature values of each feature item in the features corresponding to the service data for a risk control model applied in the preset service by a self-attention algorithm comprises: obtaining a target feature with abnormal feature values in the features corresponding to the service data, and obtaining a self-attention algorithm previously set for the target feature according to the target feature; generating, based on the feature value abnormality conditions in the features corresponding to the service data and the features without abnormal feature values in the features corresponding to the service data, model feature values of each feature item in the features corresponding to the service data for a risk control model applied in the preset service by the obtained self-attention algorithm.

4. The method of claim 2, wherein the self-attention algorithm is used to generate a model feature value of each feature item in the feature corresponding to the business data for a risk prevention and control model in the preset business based on the feature value abnormality in the feature corresponding to the business data and the feature with no feature value abnormality in the feature corresponding to the business data, comprising: obtaining a target feature with feature value abnormality in the feature corresponding to the business data, and classifying the target feature to obtain a category corresponding to each target feature, and obtaining a self-attention algorithm corresponding to each obtained category; and generating a model feature value of each feature item in the feature corresponding to the business data for a risk prevention and control model in the preset business based on the feature value abnormality in the feature corresponding to the business data, the category corresponding to each target feature, and the feature with no feature value abnormality in the feature corresponding to the business data, respectively through the self-attention algorithm corresponding to each obtained category.

5. The method of any one of claims 1-4, wherein the feature value abnormality comprises a missing feature value.

6. The method of claim 5, further comprising: determining whether the prediction result can make the risk prevention and control model converge according to a prediction result corresponding to the business data; and if not, obtaining business data of the preset business to continue model training of the risk prevention and control model until the risk prevention and control model converges, obtaining a trained risk prevention and control model, and deploying the trained risk prevention and control model in the preset business.

7. The method of claim 5, wherein before obtaining the business data of the preset business, the method further comprises: obtaining a deployment instruction of the risk prevention and control model, the risk prevention and control model being a model obtained through model training of business data of a preset business; and obtaining the risk prevention and control model based on the deployment instruction, and deploying the risk prevention and control model in the preset business.

8. A data processing apparatus, comprising: a data obtaining module configured to obtain business data of a preset business, the preset business being a payment business, and the business data comprising payment time, payment location, payment amount, account information of a payment party, and account information of a receiving party; and a feature extraction module configured to extract features of the business data to obtain a feature corresponding to the business data, the feature corresponding to the business data comprising a feature item and a feature value corresponding to the feature item, and the feature corresponding to the business data comprising a feature with feature value abnormality. ​ ​ ​ ​ ​ ​ ​ ​ a feature processing module, configured to generate, for each feature item in the features corresponding to the business data, a model feature value of a risk prevention and control model applied in the preset business based on feature value abnormality in the features corresponding to the business data and features with no feature value abnormality in the features corresponding to the business data, the feature value abnormality including what form of feature value abnormality, an influence degree of a feature with feature value abnormality on other features, an association between the feature with feature value abnormality and other features, and a feature in a feature set with a probability of feature value abnormality greater than a preset probability threshold; a prediction module, configured to input each feature item in the features corresponding to the business data and the corresponding model feature value into the risk prevention and control model to obtain a prediction result corresponding to the business data.

9. A data processing device, comprising: a processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to: obtain business data of a preset business, the preset business being a payment business, the business data including payment time, location, payment amount, account information of a payment party, and account information of a receiving party; perform feature extraction on the business data to obtain features corresponding to the business data, the features corresponding to the business data including feature items and feature values corresponding to the feature items, and the features corresponding to the business data including features with feature value abnormality; generate, for each feature item in the features corresponding to the business data, a model feature value of a risk prevention and control model applied in the preset business based on feature value abnormality in the features corresponding to the business data and features with no feature value abnormality in the features corresponding to the business data, the feature value abnormality including what form of feature value abnormality, an influence degree of a feature with feature value abnormality on other features, an association between the feature with feature value abnormality and other features, and a feature in a feature set with a probability of feature value abnormality greater than a preset probability threshold; input each feature item in the features corresponding to the business data and the corresponding model feature value into the risk prevention and control model to obtain a prediction result corresponding to the business data.

10. A storage medium for storing computer executable instructions, the executable instructions, when executed by a processor, implementing the following processes: obtaining business data of a preset business, the preset business being a payment business, the business data including payment time, location, payment amount, account information of a payment party, and account information of a receiving party; performing feature extraction on the business data to obtain features corresponding to the business data, the features corresponding to the business data including feature items and feature values corresponding to the feature items, and the features corresponding to the business data including features with feature value abnormality; generate, for each feature item in the features corresponding to the business data, a model feature value of a risk prevention and control model applied to the pre-set business, based on an abnormal feature value situation of the features corresponding to the business data and a feature with no abnormal feature value in the features corresponding to the business data, the abnormal feature value situation including what form of feature value abnormality, an influence degree of the feature value abnormality on other features, a correlation between the feature value abnormality and other features, and a feature in a feature set with a probability of feature value abnormality greater than a pre-set probability threshold; input each feature item in the features corresponding to the business data and the corresponding model feature value into the risk prevention and control model to obtain a prediction result corresponding to the business data.

Images