A method, apparatus, device, and computer storage medium for user behavior recognition.

Abstract

This application discloses a method, apparatus, device, and computer storage medium for user behavior recognition. The method includes: acquiring user behavior data and user historical state information; extracting feature data related to the user state information from the behavior data to obtain first target feature data; determining first state information corresponding to the first target feature data based on the correspondence between the feature data and state information; calculating a behavior index value from the first state information and historical state information; and determining the user behavior corresponding to the user behavior data as the target user behavior based on the behavior index value. According to the embodiments of this application, the accuracy of recognizing user operation behaviors is improved.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and in particular relates to a method, apparatus, device and computer storage medium for user behavior recognition. Background Technology

[0002] Typically, risk control systems are used to monitor users' online behavior in order to identify whether the user's behavior is normal or abnormal.

[0003] Currently, existing detection methods only identify a portion of a user's actions within a short period, judging whether the actions are normal based on the identification results. However, users can achieve the same purpose as abnormal actions through multiple normal operations. Identifying only a portion of the user's actions cannot distinguish this type of operation from abnormal operations, resulting in low accuracy in user behavior identification. Furthermore, using the above detection methods, even accidental user actions will be identified as abnormal, leading to poor accuracy in identifying user behavior. Summary of the Invention

[0004] This application provides a method, apparatus, device, and computer storage medium for user behavior recognition, which can improve the accuracy of recognizing user operation behavior.

[0005] In a first aspect, embodiments of this application provide a method for user behavior recognition, the method comprising:

[0006] Obtain user behavior data and user historical status information;

[0007] Extract feature data related to historical state information from behavioral data to obtain the first target feature data;

[0008] Based on the correspondence between feature data and state information, determine the first state information corresponding to the first target feature data;

[0009] Behavioral index values ​​are calculated from the first state information and historical state information;

[0010] The user behavior corresponding to the user behavior data is determined based on the behavioral indicator value as the target user behavior.

[0011] In some embodiments, the behavioral indicator value is calculated from the first state information and the historical state information, including:

[0012] Calculate the first state information and historical state data to obtain multiple target state information of the user, and the target state information represents the user's real state information;

[0013] The behavioral index value is obtained by weighted summation of multiple target state data.

[0014] In some embodiments, there are multiple behavioral indicator values; after calculating the behavioral indicator values ​​from the first state information and historical state information, the method further includes:

[0015] Feature extraction is performed on behavioral indicator values ​​and behavioral data to obtain the second target feature data;

[0016] Based on the correspondence between the second target feature data and the processing method for the user, determine the processing method corresponding to the second target feature data;

[0017] The user was processed according to the established procedures.

[0018] In some embodiments, feature extraction is performed on behavioral indicator values ​​and behavioral data to obtain second target feature data, including:

[0019] Based on behavioral data, determine behavioral indicator values ​​that meet preset conditions for correlation with behavioral data from the behavioral indicator values ​​to obtain multiple target behavioral indicator values;

[0020] The credibility index value is obtained by weighted summation of multiple target index values.

[0021] The credibility index value is used as the second target feature data.

[0022] In some embodiments, after calculating the first state information and historical state data to obtain multiple target state information of the user, the method further includes:

[0023] Update historical state data with multiple target state information.

[0024] In some embodiments, after acquiring user behavior data and historical state data, the method further includes:

[0025] Classify the behavioral data to obtain the behavioral types of the data;

[0026] The forgetting parameters of the behavioral data are determined based on the behavior type; the forgetting parameters include multiple parameter values; the historical state data includes multiple first sub-data that correspond one-to-one with the multiple parameter values; the parameter values ​​represent the degree of correlation between the behavioral features corresponding to the behavior type and the corresponding first sub-data.

[0027] Calculations are performed on the first state information and historical state data to obtain multiple target state information for the user, including:

[0028] For each first sub-data in a plurality of first sub-data, calculate the product of the first sub-data and its corresponding parameter value to obtain the target sub-data corresponding to the first sub-data;

[0029] The third sub-data in the first state information is weighted and averaged with the corresponding target sub-data in the multiple target sub-data to obtain multiple target state data.

[0030] In some embodiments, before determining the forgetting parameter of the behavioral data based on the behavior type, the method further includes:

[0031] Obtain risk control requirements data corresponding to the behavior type;

[0032] Multiple second sub-data points are randomly selected from the risk control requirements data, and this selection is repeated multiple times to obtain multiple datasets.

[0033] For each dataset, the intermediate forgetting parameter is calculated for multiple second sub-data points in each dataset. The intermediate forgetting parameter represents the degree of correlation between the behavioral features corresponding to multiple second sub-data points and the corresponding first sub-data points. The intermediate forgetting parameter includes multiple intermediate forgetting parameter values.

[0034] Obtain multiple weight sets; each weight set includes a weight value corresponding to each intermediate forgetting parameter value in each intermediate forgetting parameter.

[0035] For each weight set, the target parameter value corresponding to the weight set is obtained by weighting and summing multiple intermediate forgotten parameter values ​​using the weight values.

[0036] The forgetting parameters are obtained based on the target parameter values ​​corresponding to each weight set.

[0037] In some embodiments, intermediate forgetting parameters related to the second sub-data in each dataset are calculated for multiple second sub-data in each dataset, including:

[0038] For each dataset, multiple second sub-data points are weighted and summed to obtain the intermediate forgetting parameters related to the second sub-data points in each dataset.

[0039] Secondly, embodiments of this application provide a user behavior recognition device, comprising:

[0040] The acquisition module is used to acquire user behavior data and user historical status information;

[0041] The extraction module is used to extract feature data related to user state information from behavioral data to obtain the first target feature data;

[0042] The first determining module is used to determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information.

[0043] The calculation module is used to calculate behavioral indicator values ​​from the first state information and historical state information;

[0044] The second determination module is used to determine the user behavior corresponding to the user behavior data as the target user behavior based on the behavior indicator value.

[0045] Thirdly, embodiments of this application provide an electronic device, the device including: a processor and a memory storing computer program instructions;

[0046] A method for user behavior recognition as described in any of the first aspects when the processor executes computer program instructions.

[0047] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the user behavior recognition method as described in any of the first aspects.

[0048] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform a user behavior recognition method as described in any of the first aspects.

[0049] The user behavior recognition method, apparatus, device, and computer storage medium of this application embodiment can extract first target feature data related to the user's historical state information from the user's behavior data, determine the first state information corresponding to the first target feature data according to the correspondence between the feature data and the state information, calculate the behavior index value by calculating the first state information and the historical state information, and determine the user behavior corresponding to the user behavior data as the target user behavior based on the behavior index value. Since the behavior index value is obtained by the first state information and the historical state information corresponding to the behavior data, it can consider not only the user's short-term behavior data but also the user's historical state information when recognizing user behavior, that is, it considers the user's long-term historical operation behavior, and thus comprehensively judge whether the user behavior has risks, thereby improving the accuracy of recognizing user operation behavior. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating a user behavior recognition method provided in one embodiment of this application;

[0052] Figure 2This is a flowchart illustrating a user behavior recognition method provided in another embodiment of this application;

[0053] Figure 3 This is a flowchart illustrating a user behavior recognition method provided in another embodiment of this application;

[0054] Figure 4 This is a schematic diagram of the architecture corresponding to an embodiment of the user identification method provided in this application.

[0055] Figure 5 This is a schematic diagram of the structure of a user behavior recognition device provided in one embodiment of this application;

[0056] Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0057] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0058] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0059] Furthermore, the acquisition, storage, use, and processing of data in this application's technical solution all comply with relevant national laws and regulations.

[0060] As mentioned above, existing technologies that use risk control systems to monitor users' online behavior suffer from poor accuracy in identifying user behavior.

[0061] To address the aforementioned issues, this application proposes a method, apparatus, device, and computer storage medium for user behavior recognition. This method extracts first target feature data related to the user's historical state information from the user's behavior data. Based on the correspondence between the feature data and state information, it determines the first state information corresponding to the first target feature data. It then calculates a behavior index value from the first state information and the historical state information. Based on the behavior index value, it determines the user behavior corresponding to the user behavior data as the target user behavior. Since the behavior index value is obtained through the first state information and historical state information corresponding to the behavior data, it considers not only the user's short-term behavior data but also the user's historical state information, i.e., the user's long-term historical operational behavior, thus comprehensively judging whether the user behavior poses a risk and improving the accuracy of user behavior recognition.

[0062] The user behavior recognition method provided in the embodiments of this application will be introduced first below.

[0063] Figure 1 This is a flowchart illustrating a user behavior recognition method provided in one embodiment of this application. Figure 1 As shown, the user behavior recognition method provided in this application embodiment includes the following steps: S101-S105.

[0064] S101: Obtain user behavior data and user historical status information.

[0065] In one embodiment of this application, historical status information can be status information obtained based on the user's historical behavior, such as the user's historical risk score or credit rating.

[0066] In another embodiment of this application, the aforementioned behavioral data may be the user's online or offline operational behavior data. Furthermore, the behavioral data may be acquired in real time or acquired at a preset frequency. Behavioral data may include incorrect user account or password input, the number of times the user transferred funds within a certain period, and changes in the terminal device used by the user.

[0067] S102: Extract feature data related to historical state information from the behavioral data to obtain the first target feature data.

[0068] Feature data related to historical state information can be extracted from behavioral data, and this feature data can be used as the first target feature data.

[0069] In one embodiment, a pre-trained neural network model can be used to extract features from the aforementioned behavioral data to obtain first target feature data related to historical states.

[0070] In one example, behavioral data includes the fact that a user made 10 transfers within 10 minutes, and historical status information includes risk level status information corresponding to the user's historical transfer frequency. Therefore, the first target feature data obtained by feature extraction of the user status can be the user's transfer frequency, which is 1 minute / transfer.

[0071] It should be noted that the above neural network model can be any of the following models: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Recurrent Neural Network (RNN), or Transformer model.

[0072] S103: Determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information.

[0073] It should be noted that since the above feature data is feature data extracted from user behavior data and associated with historical state information, there is a correspondence between feature data and state information. It can be understood that there is also a correspondence between feature data and historical state information.

[0074] The first state information corresponding to the first target feature data can be determined according to the correspondence between feature data and historical state information.

[0075] S104: Calculate the behavior index value by combining the first state information and the historical state information.

[0076] In one embodiment, the first state information can be obtained based on the user's behavior data. In order to more accurately identify user behavior, it is necessary to consider the user's historical behavior, that is, to calculate based on the first state information and the historical state information to obtain the behavior index value.

[0077] In one embodiment, a behavioral indicator value is calculated from the first state information and historical state information. The calculation method described above can be set according to the relevant requirements corresponding to actual user behavior.

[0078] For example, the first target feature data is a transfer frequency of 1 minute / time. The correspondence between the feature data and the status information is as follows: if the transfer frequency is greater than 10 minutes / time, the weight value corresponding to the level 1 risk status in the status information is 0.4; if the transfer frequency is less than or equal to 10 minutes / time, the weight value corresponding to the level 2 risk status in the status information is 0.6. Therefore, the first target feature data is determined to correspond to the level 2 risk status with a weight of 0.6, and the historical status information is the level 1 risk status with a corresponding weight value of 0.4. Thus, the behavioral indicator value = (1 x 0.4 + 2 x 0.6) / 2 = 0.8.

[0079] In another embodiment, nonlinear operations can be performed on the first state information and historical state information to obtain behavioral index values. The aforementioned nonlinear operations can be set according to actual needs.

[0080] S105: Determine the user behavior corresponding to the user behavior data based on the behavioral indicator value as the target user behavior.

[0081] In one embodiment, the user behavior corresponding to the user behavior data can be determined as the target user behavior based on the level of the behavior indicator value.

[0082] As an example, the target user behavior can be high-risk, medium-risk, low-risk, or risk-free.

[0083] The user behavior recognition method provided in this application extracts first target feature data related to the user's historical state information from the user's behavior data. Based on the correspondence between the feature data and the state information, it determines the first state information corresponding to the first target feature data. It calculates a behavior index value by combining the first state information and the historical state information. Based on the behavior index value, it determines the user behavior corresponding to the user behavior data as the target user behavior. Since the behavior index value is obtained through the first state information and historical state information corresponding to the behavior data, it can consider not only the user's short-term behavior data but also the user's historical state information when recognizing user behavior, that is, it considers the user's long-term historical operation behavior. This allows for a comprehensive judgment on whether the user behavior is risky, thus improving the accuracy of recognizing user operation behavior.

[0084] This application proposes another method for user behavior recognition; please refer to [link to relevant documentation]. Figure 2 .like Figure 2 As shown, the user behavior recognition method includes:

[0085] S201: Obtain user behavior data and user historical status information.

[0086] S202: Extract feature data related to historical state information from the behavioral data to obtain the first target feature data.

[0087] S203: Determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information.

[0088] Steps S201 to S203 in this embodiment are the same as S201 to S203 in the embodiments of this application. For the sake of brevity, they will not be described in detail here. For detailed information, please refer to the description in the embodiments of this application.

[0089] S204: Calculate the first state information and historical state data to obtain multiple target state information of the user.

[0090] Among them, the target state information represents the user's actual state information.

[0091] It should be noted that the first state information may include multiple sub-information, and the historical state data may include the same number of sub-data as the sub-information, with a one-to-one correspondence between the sub-information and the sub-data.

[0092] In one embodiment, for each of the multiple pieces of sub-information, a weighted average of the sub-information and its corresponding sub-data can be calculated to obtain the target state information corresponding to that sub-information. This results in multiple target state information sets.

[0093] S205: The behavioral indicator value is obtained by weighted summation of multiple target state data.

[0094] In one embodiment, multiple target state data can be weighted and summed according to corresponding weight values ​​to obtain the aforementioned behavioral indicator value. These weight values ​​can be pre-set or obtained through non-linear calculations based on the user behavior data according to a preset calculation method.

[0095] In some embodiments of this application, there are multiple behavioral indicator values. After S205, the method further includes:

[0096] Feature extraction is performed on behavioral indicator values ​​and behavioral data to obtain the second target feature data;

[0097] Based on the correspondence between the second target feature data and the processing method for the user, determine the processing method corresponding to the second target feature data;

[0098] The user was processed according to the established procedures.

[0099] In one embodiment, a pre-trained convolutional neural network can be used to extract features from behavioral index values ​​and behavioral data to obtain second target feature data. The second target feature data can be a high-dimensional feature vector.

[0100] After obtaining the second target feature data, the processing method corresponding to the second target feature data can be determined based on the correspondence between the second target feature data and the processing method for the user.

[0101] Here, the correspondence between the second target feature data and the processing method for the user can be pre-set, or it can be predicted by a pre-trained convolutional neural network. That is, the convolutional neural network is trained in advance using training data so that it learns to accurately predict the correspondence between the second target feature data and the processing method. The trained convolutional neural network is then used to predict the processing method corresponding to the second target feature data.

[0102] In addition, as an example, the approach could be to blacklist the user's account, disable the user's permission to conduct related business through offline counters, or reduce the user's maximum transfer limit.

[0103] Thus, by establishing the correspondence between the second target feature data and the processing methods for users, the processing method corresponding to the second target feature data is determined, and the user is processed according to the processing method. In other words, corresponding measures are taken for the user based on the processing method to reduce the risks caused by user behavior.

[0104] In some embodiments of this application, there may be multiple behavioral indicator values. The above-mentioned feature extraction of behavioral indicator values ​​and behavioral data to obtain second target feature data may specifically include:

[0105] Based on behavioral data, determine behavioral indicator values ​​that meet preset conditions for correlation with behavioral data from the behavioral indicator values ​​to obtain multiple target behavioral indicator values;

[0106] The credibility index value is obtained by weighted summation of multiple target index values.

[0107] The credibility index value is used as the second target feature data.

[0108] It should be noted that the above-mentioned behavioral indicator values ​​can all be obtained through the specific implementation methods proposed in any embodiment of this application for calculating behavioral indicator values ​​from first state information and historical state information.

[0109] In one embodiment, a pre-trained convolutional neural network can be used to extract behavioral indicator values ​​from multiple behavioral indicator values ​​that have a higher correlation with the user's behavioral data than a preset threshold, and these behavioral indicator values ​​can be used as target behavioral indicator values. It is understood that there can be multiple target behavioral indicator values.

[0110] As an example, the aforementioned behavioral indicator values ​​may include the user's credit score, the user's behavior reasonableness score, and the risk score of the device used by the user. If the user's behavior is due to an incorrect account password, the convolutional neural network can determine the behavioral indicator values ​​most relevant to the aforementioned user behavior as the user's behavior reasonableness score and user credit score. The user's behavior reasonableness score and user credit score are then weighted and summed to obtain the credibility indicator value.

[0111] In one embodiment, after obtaining multiple target behavior indicator values, the multiple target behavior indicator values ​​can be weighted and summed to obtain a credibility indicator value, which is then used as the second target feature data. Specifically, the multiple target behavior indicator values ​​can be weighted and summed according to the weight value corresponding to each target behavior indicator, and the weight value can be preset.

[0112] S206: Determine the user behavior corresponding to the user behavior data based on the behavioral indicator value as the target user behavior.

[0113] In this embodiment, S206 is the same as S205 in the embodiment of this application. For the sake of brevity, it will not be described in detail here. For detailed information, please refer to the description in the embodiment of this application.

[0114] S207: Update historical state data to multiple target state information.

[0115] In one embodiment, after each user behavior is identified, the historical state data can be updated based on the latest determined user target state information, i.e., the user's actual state information. Thus, the updated historical state data is determined based on historical user behavior and currently collected user behavior. The updated historical state data is used for identification in the next user behavior identification, thereby realizing the identification of the latest collected user behavior based on long-term user behavior data, further improving the accuracy of identifying user operation behavior.

[0116] In another embodiment, if the number of sub-data in the historical state data is less than the number of sub-data in the multiple target state information, the multiple target state information can be input into a pre-trained convolutional neural network. The convolutional neural network extracts features from the multiple target state information to obtain state information corresponding to the sub-data in the historical state data, and updates the historical state data accordingly.

[0117] The above describes a user behavior recognition method provided in this application embodiment. This method calculates multiple target state information of the user based on first state information and historical state data. The target state information represents the user's true state information. The multiple target state data are weighted and summed to obtain a behavior indicator value. The behavior indicator value is determined based on the user's true state information, which improves the accuracy of determining the behavior indicator value. In addition, the user's processing method is determined based on the behavior indicator value, so that corresponding measures can be taken for the user according to the processing method, reducing the risks that user behavior may cause. Furthermore, by updating historical state information, the accuracy of recognizing user operation behavior is further improved.

[0118] This application proposes another method for user behavior recognition; please refer to [link to relevant documentation]. Figure 3 .like Figure 3 As shown, the user behavior recognition method includes:

[0119] S301: Obtain user behavior data and user historical status information.

[0120] In this embodiment, S301 is the same as S101 in the embodiment of this application. For the sake of brevity, it will not be described in detail here. For details, please refer to the description in the embodiment of this application.

[0121] S302: Classify the behavioral data to obtain the behavioral type of the behavioral data.

[0122] In some embodiments of this application, behavioral data includes the number of times a user makes transfers within a preset time period, and S302 may specifically include:

[0123] Input the number of transfers into the pre-trained decision tree model;

[0124] Based on the preset threshold classification conditions for the number of transfers corresponding to the multi-level non-leaf nodes in the decision tree model, the number of transfers is classified level by level until the leaf node corresponding to the last level non-leaf node is obtained.

[0125] Use the behavior type corresponding to the leaf node as the behavior type for the number of transfers.

[0126] It's understandable that decision tree models can be pre-trained based on training set data. These models can be used to classify transaction counts and determine the corresponding behavioral types.

[0127] In some embodiments of this application, the decision tree model can be any one of the following: simple decision tree algorithm, ID3 tree model, C4.5 tree model, and classification and regression CART tree model.

[0128] S303: Determine the forgetting parameters of behavioral data based on the behavior type.

[0129] In one embodiment, the forgetting parameter includes multiple parameter values; the historical state data includes multiple first sub-data that correspond one-to-one with the multiple parameter values; the parameter value represents the degree of correlation between the behavioral feature corresponding to the behavioral type and the corresponding first sub-data.

[0130] It should be noted that the forgetting parameter can be a high-dimensional feature vector, and the forgetting parameter can include multiple parameter values, each of which can correspond to a feature associated with the user's behavior type. Historical state data can include multiple first sub-data points, each of which corresponds to a parameter value, meaning there is a one-to-one correspondence between parameter values ​​and first sub-data points.

[0131] In some embodiments of this application, prior to S303, the method may further include:

[0132] Obtain risk control requirements data corresponding to the behavior type;

[0133] Multiple second sub-data points are randomly selected from the risk control requirements data, and this selection is repeated multiple times to obtain multiple datasets.

[0134] For each dataset, the intermediate forgetting parameter is calculated for multiple second sub-data points in each dataset. The intermediate forgetting parameter represents the degree of correlation between the behavioral features corresponding to multiple second sub-data points and the corresponding first sub-data points. The intermediate forgetting parameter includes multiple intermediate forgetting parameter values.

[0135] Obtain multiple weight sets; each weight set includes a weight value corresponding to each intermediate forgetting parameter value in each intermediate forgetting parameter.

[0136] For each weight set, the target parameter value corresponding to the weight set is obtained by weighting and summing multiple intermediate forgotten parameter values ​​using the weight values.

[0137] The forgetting parameters are obtained based on the target parameter values ​​corresponding to each weight set.

[0138] It should be noted that the aforementioned risk control requirements data can be pre-set based on actual needs. Risk control requirements data can be text data or structured data that can be directly understood by computers. Furthermore, risk control requirements can be risk control strategies formulated by the risk control department, or ratings of different emails based on business needs.

[0139] Before determining the forgetting parameters of behavioral data based on the behavior type, risk control requirement data corresponding to that behavior type can be obtained first.

[0140] If the risk control requirements data is text data, the text data can be preprocessed first to extract the structured risk control requirements data.

[0141] As an example, the risk control requirement data includes M second sub-data. N second sub-data can be randomly selected from the M second sub-data, and this selection can be repeated multiple times to obtain multiple datasets consisting of N second sub-data, where M is an integer greater than 1 and N is an integer greater than 0 and less than M.

[0142] For each of the above datasets, the intermediate forgetting parameters corresponding to each dataset are calculated by analyzing multiple second sub-data points within the dataset.

[0143] It should be noted that the above calculation of multiple second sub-data points can be a non-linear operation. The intermediate forgetting parameter represents the degree of correlation between the behavioral features corresponding to the multiple second sub-data points and the corresponding first sub-data points, and the intermediate forgetting parameter includes multiple intermediate forgetting parameter values.

[0144] In one embodiment, multiple weight sets can be obtained, the number of which is the same as the number of parameter values ​​in the forgetting parameters to be determined. Each weight set includes a weight value corresponding to each intermediate forgotten parameter value in each intermediate forgotten parameter. For each weight set, the multiple intermediate forgotten parameter values ​​are weighted and summed using the weight values ​​to obtain the target parameter value corresponding to the weight set. The target parameter value corresponding to each weight set is then used as the parameter value in the forgetting parameters to obtain the forgetting parameters.

[0145] It should be noted that the aforementioned weight set can be pre-set. The method for generating the forgetting parameters can be implemented using a regression random forest model. The aforementioned multiple weight sets can be obtained by pre-training a regression random forest model.

[0146] In another embodiment, any one of linear regression, multinomial regression, support vector regression and artificial neural network regression, decision tree regression, lasso regression, ridge regression, and ElasticNet regression can be used to determine the forgetting parameter.

[0147] In some embodiments of this application, after obtaining the aforementioned forgetting parameters, regularization processing can be performed on the forgetting parameters, thereby reducing the amount of data processing and enabling the forgetting parameter values ​​to be processed by the neural network proposed in the embodiments of this application. The regularization processing can employ any one of the following functions: sigmoid function, hyperbolic tangent tanh function, linear rectified ReLU function, modified linear unit LeakyReLU function, parameter modified linear unit PReLU function, and exponential linear unit ELU function.

[0148] In some embodiments of this application, the above-mentioned calculation of intermediate forgetting parameters related to the second sub-data in each dataset for multiple second sub-data in each dataset may specifically include:

[0149] For each dataset, multiple second sub-data points are weighted and summed to obtain the intermediate forgetting parameters related to the second sub-data points in each dataset.

[0150] In one embodiment, the above steps are implemented using a pre-trained regression random forest. For the decision tree in the random regression forest, the data processing method in the root node of the decision tree can be used to combine multiple second sub-data into multiple datasets according to a preset combination method. The data in each dataset is weighted and summed to obtain a weighted sum value. The child node that matches the weighted sum value is found, that is, the weighted sum value satisfies the preset condition corresponding to the child node, thereby obtaining the parameter value stored in the leaf node corresponding to the child node. The parameter values ​​corresponding to the above multiple datasets are used as the parameter values ​​of the intermediate forgetting parameters, thereby obtaining the intermediate forgetting parameters related to the second sub-data.

[0151] S304: Extract feature data related to historical state information from the behavioral data to obtain the first target feature data.

[0152] S305: Determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information.

[0153] Steps S304 to S305 in this embodiment are the same as S102 to S103 in the embodiments of this application. For the sake of brevity, they will not be described in detail here. For detailed information, please refer to the description in the embodiments of this application.

[0154] S306: For each of the multiple first sub-data, calculate the product of the first sub-data and its corresponding parameter value to obtain the target sub-data corresponding to the first sub-data.

[0155] It is understandable that, since the above parameter values ​​represent the degree of correlation between the behavioral characteristics corresponding to the behavior type and the corresponding first sub-data, the first sub-data can be weighted according to different degrees of correlation. That is, the historical state data is processed according to the degree of correlation to obtain historical state data that is more consistent with the actual situation of user behavior data, thereby improving the accuracy of identifying user behavior.

[0156] S307: Take a weighted average of each third sub-data in the first state information and the corresponding target sub-data in the multiple target sub-data to obtain multiple target state data.

[0157] In one embodiment, the first state information may include multiple third sub-data of different dimensions, each of which corresponds to a target sub-data. For each third sub-data, a weighted average of the third sub-data and its corresponding target sub-data can be performed to obtain the target state data corresponding to the third sub-data, thereby obtaining the aforementioned multiple target state data.

[0158] S308: The behavioral index value is obtained by weighted summation of multiple target state data.

[0159] In this embodiment, S308 is the same as S205 in the embodiment of this application. For the sake of brevity, it will not be described in detail here. For detailed information, please refer to the description in the embodiment of this application.

[0160] S309: Determine the user behavior corresponding to the user behavior data based on the behavioral indicator value as the target user behavior.

[0161] In this embodiment, S309 is the same as S105 in the embodiment of this application. For the sake of brevity, it will not be described in detail here. For details, please refer to the description in the embodiment of this application.

[0162] The above is a user behavior recognition method provided in the embodiments of this application. In the process of judging user behavior, the method adds a forgetting parameter to represent the degree of correlation between the behavior feature corresponding to the behavior type and the corresponding first sub-data. According to the degree of correlation between the user behavior type and the historical state data, the first sub-data in the historical state data is adjusted to improve the accuracy of user behavior recognition.

[0163] Figure 4 This is a schematic diagram of the architecture corresponding to an embodiment of the user identification method provided in this application.

[0164] like Figure 4 As shown, user behavior data is collected in real time and input into a user behavior classification system (which can be the decision tree model proposed in this embodiment). The user behavior classification system classifies the behavior data to obtain the type of behavior data, determines the forgetting parameter corresponding to the type, and inputs the forgetting parameter, behavior data, and user state information (i.e., the historical state information proposed in this embodiment) into a neural network. The neural network predicts and outputs the real state information of multiple users, including outputting the user's behavior score (i.e., the behavior index value proposed in this embodiment). The user behavior corresponding to the user behavior data is determined based on the behavior score.

[0165] Alternatively, user behavior scores can be input into a basic risk control system (which may be the convolutional neural network proposed in this application). The basic risk control system determines a processing plan corresponding to the user behavior based on the user behavior scores, and processes the user according to the processing plan.

[0166] Furthermore, when the real state information does not match the historical state information, the real state information can be input into the user state update system (which can use a pre-trained convolutional neural network). The user state update system can extract the new user state information corresponding to the historical state information from the real state information and update the historical state information to the new user state information, so that the next time the newly collected user behavior is identified, it can be identified based on the updated historical state information.

[0167] Based on the user behavior recognition method provided in the above embodiments, combined with the appendix Figure 5 This application describes specific implementations of the user behavior recognition device provided in its embodiments.

[0168] See Figure 5 This is a schematic diagram of the structure of a user behavior recognition device 500 provided in one embodiment of this application. The user behavior recognition device 500 includes:

[0169] The acquisition module 501 is used to acquire user behavior data and user historical status information;

[0170] Extraction module 502 is used to extract feature data related to user state information from behavioral data to obtain first target feature data;

[0171] The determining module 503 is used to determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information;

[0172] The first calculation module 504 is used to calculate the behavior index value from the first state information and the historical state information;

[0173] The determination module 503 is also used to determine the user behavior corresponding to the user behavior data as the target user behavior based on the behavior indicator value.

[0174] The training device for the bearing fault diagnosis model provided in this application extracts first target feature data related to the user's historical state information from the user's behavior data. Based on the correspondence between the feature data and the state information, it determines the first state information corresponding to the first target feature data. It calculates the behavior index value by calculating the first state information and the historical state information. Based on the behavior index value, it determines the user behavior corresponding to the user behavior data as the target user behavior. Since the behavior index value is obtained by the first state information and the historical state information corresponding to the behavior data, it can consider not only the user's short-term behavior data but also the user's historical state information when identifying user behavior, that is, it considers the user's long-term historical operation behavior. In this way, it can comprehensively judge whether the user behavior has risks and improve the accuracy of identifying user operation behavior.

[0175] As one implementation of this application, the first determining module 503 described above may include:

[0176] The calculation submodule is used to calculate the first state information and historical state data to obtain multiple target state information of the user. The target state information represents the user's real state information.

[0177] The weighted summation submodule is used to perform weighted summation on multiple target state data to obtain behavioral indicator values.

[0178] As one implementation of this application, there are multiple behavioral indicator values. To reduce the risks that user behavior may cause, the above-mentioned device may further include:

[0179] The extraction module is used to extract features from behavioral indicator values ​​and behavioral data to obtain the second target feature data.

[0180] The determination module is also used to determine the processing method corresponding to the second target feature data based on the correspondence between the second target feature data and the processing method for the user;

[0181] The processing module is used to process users according to the processing method.

[0182] As one implementation of this application, the extraction module described above may include:

[0183] The determination submodule is used to determine the behavioral indicator values ​​that meet the preset conditions of correlation with the behavioral data from the behavioral indicator values ​​based on the behavioral data, and to obtain multiple target behavioral indicator values;

[0184] The weighted summation submodule is also used to sum multiple target index values ​​in a weighted manner to obtain a credibility index value;

[0185] The configuration submodule is used to use the credibility index value as the second target feature data.

[0186] As one implementation of this application, in order to further improve the accuracy of recognizing user operation behavior, the above-mentioned device may include:

[0187] The update module is used to update historical state data with multiple target state information.

[0188] As one implementation of this application, to further improve the accuracy of recognizing user operation behaviors, the above-mentioned device may further include:

[0189] The classification module is used to classify behavioral data to obtain the behavioral type of the data;

[0190] The determination module 503 is also used to determine the forgetting parameters of the behavior data according to the behavior type; the forgetting parameters include multiple parameter values; the historical state data includes multiple first sub-data that correspond one-to-one with the multiple parameter values; the parameter values ​​represent the degree of correlation between the behavior features corresponding to the behavior type and the corresponding first sub-data;

[0191] Specifically, the calculation submodule can be used to: calculate the product of the first sub-data and its corresponding parameter value for each of the multiple first sub-data to obtain the target sub-data corresponding to the first sub-data; and perform a weighted average of each third sub-data in the first state information and the corresponding target sub-data in the multiple target sub-data to obtain multiple target state data.

[0192] As one implementation of this application, the above-mentioned apparatus may further include:

[0193] The acquisition module 501 is also used to acquire risk control requirement data corresponding to the behavior type;

[0194] The selection module is used to randomly select multiple second sub-data from the risk control requirement data, and to randomly select multiple times to obtain multiple datasets;

[0195] The calculation module is also used to calculate the intermediate forgetting parameters for each dataset by calculating the multiple second sub-data in each dataset. The intermediate forgetting parameters represent the degree of correlation between the behavioral features corresponding to the multiple second sub-data and the corresponding first sub-data. The intermediate forgetting parameters include multiple intermediate forgetting parameter values.

[0196] The acquisition module 501 is also used to acquire multiple weight sets; the weight sets include weight values ​​corresponding to each intermediate forgetting parameter value in each intermediate forgetting parameter;

[0197] The weighted summation module is used to sum multiple intermediate forgotten parameter values ​​using the weight values ​​for each weight set, so as to obtain the target parameter value corresponding to the weight set.

[0198] The determination module 505 is also used to obtain the forgetting parameters based on the target parameter values ​​corresponding to each weight set.

[0199] As one implementation of this application, the determining module 503 may include:

[0200] The weighted summation submodule is also used to perform weighted summation on multiple second subdata for each dataset to obtain intermediate forgetting parameters related to the second subdata in each dataset.

[0201] Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application.

[0202] Electronic device 600 may include processor 601 and memory 602 storing computer program instructions.

[0203] Specifically, the processor 601 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0204] Memory 602 may include a large-capacity memory for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 602 may include removable or non-removable (or fixed) media. Where appropriate, memory 602 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 602 is non-volatile solid-state memory. Memory 602 may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical / tangible memory storage devices. Thus, typically, memory 602 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the user behavior recognition methods in the above embodiments.

[0205] The processor 601 implements any of the user behavior recognition methods in the above embodiments by reading and executing computer program instructions stored in the memory 602.

[0206] In one example, electronic device 600 may further include communication interface 603 and bus 610. For example, Figure 6 As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 610 and complete communication with each other.

[0207] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0208] Bus 610 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 610 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0209] Furthermore, in conjunction with the user behavior recognition methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the user behavior recognition methods in the above embodiments.

[0210] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0211] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0212] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0213] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0214] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for user behavior recognition, characterized in that, include: Acquire user behavior data and the user's historical state information, wherein the historical state information is state information obtained based on the user's historical behavior, and the behavior data is the user's online or offline operation behavior data; Extract feature data related to historical state information from the behavioral data to obtain the first target feature data; Based on the correspondence between feature data and state information, determine the first state information corresponding to the first target feature data; Behavioral index values ​​are calculated from the first state information and the historical state information; Based on the behavioral indicator values, the user behavior corresponding to the user's behavioral data is determined as the target user behavior; The behavioral indicator value is calculated from the first state information and the historical state information, including: The behavioral data is classified to obtain the behavioral type of the behavioral data; The forgetting parameter of the behavior data is determined according to the behavior type; the forgetting parameter is a high-dimensional feature vector, and the forgetting parameter includes multiple parameter values; the historical state information includes multiple first sub-data that correspond one-to-one with the multiple parameter values; the parameter value represents the degree of correlation between the behavior feature corresponding to the behavior type and the corresponding first sub-data. The first state information and the historical state information are calculated based on the forgetting parameter to obtain multiple target state information of the user, wherein the target state information represents the user's true state information; The behavioral index value is obtained by weighted summation of the multiple target state information; The step of calculating multiple target state information of the user based on the forgetting parameter using the first state information and the historical state information includes: For each of the plurality of first sub-data, calculate the product of the first sub-data and its corresponding parameter value to obtain the target sub-data corresponding to the first sub-data; The weighted average of each third sub-data in the first state information and the corresponding target sub-data in the multiple target sub-data is obtained to obtain the multiple target state information.

2. The user behavior recognition method according to claim 1, characterized in that, There are multiple behavioral indicator values; After calculating the behavioral indicator value from the first state information and the historical state information, the method further includes: Feature extraction is performed on the behavioral indicator values ​​and the behavioral data to obtain the second target feature data; Based on the correspondence between the second target feature data and the processing method for the user, determine the processing method corresponding to the second target feature data; The user is processed according to the described processing method.

3. The user behavior recognition method according to claim 2, characterized in that, The step of extracting features from the behavioral indicator values ​​and the behavioral data to obtain second target feature data includes: Based on the behavioral data, determine the behavioral indicator values ​​that meet the preset conditions for correlation with the behavioral data from the behavioral indicator values ​​to obtain multiple target behavioral indicator values; The credibility index value is obtained by weighted summation of the multiple target behavior index values. The credibility index value is used as the second target feature data.

4. The user behavior recognition method according to claim 1, characterized in that, After calculating the first state information and the historical state information based on the forgetting parameter to obtain multiple target state information of the user, the method further includes: The historical state information is updated to the multiple target state information.

5. The user behavior recognition method according to claim 1, characterized in that, Before determining the forgetting parameter of the behavior data based on the behavior type, the method further includes: Obtain risk control requirement data corresponding to the behavior type; Multiple second sub-data sets are randomly selected from the risk control requirement data, and this selection is repeated multiple times to obtain multiple datasets. For each dataset, intermediate forgetting parameters are calculated for multiple second sub-data points in each dataset; the intermediate forgetting parameters represent the degree of correlation between the behavioral features corresponding to the multiple second sub-data points and the corresponding first sub-data points; the intermediate forgetting parameters include multiple intermediate forgetting parameter values. Obtain multiple weight sets; the weight sets include weight values ​​corresponding to each intermediate forgetting parameter value in each intermediate forgetting parameter. For each weight set, the weight values ​​are used to perform a weighted summation of the multiple intermediate forgotten parameter values ​​to obtain the target parameter value corresponding to the weight set; The forgetting parameters are obtained based on the target parameter values ​​corresponding to each weight set.

6. The user behavior recognition method according to claim 5, characterized in that, The step of calculating intermediate forgetting parameters related to the second sub-data in each dataset for each dataset includes: For each dataset, the multiple second sub-data are weighted and summed to obtain the intermediate forgetting parameters related to the second sub-data in each dataset.

7. A device for user behavior recognition, characterized in that, include: The acquisition module is used to acquire user behavior data and the user's historical state information. The historical state information is state information obtained based on the user's historical behavior, and the behavior data is the user's online or offline operation behavior data. The extraction module is used to extract feature data related to user state information from the behavior data to obtain the first target feature data; The determination module is used to determine the first state information corresponding to the first target feature data based on the correspondence between feature data and state information; The calculation module is used to calculate the behavioral indicator value from the first state information and the historical state information; The determining module is further configured to determine, based on the behavior indicator value, the user behavior corresponding to the user's behavior data as the target user behavior; The calculation module is specifically used for: classifying the behavioral data to obtain the behavioral type of the behavioral data; determining the forgetting parameter of the behavioral data based on the behavioral type; the forgetting parameter is a high-dimensional feature vector, and the forgetting parameter includes multiple parameter values; the historical state information includes multiple first sub-data that correspond one-to-one with the multiple parameter values; the parameter value represents the degree of correlation between the behavioral feature corresponding to the behavioral type and the corresponding first sub-data; and calculating multiple target state information of the user based on the forgetting parameter, wherein the target state information represents the user's true state information. The behavioral index value is obtained by weighted summation of the multiple target state information; For each of the plurality of first sub-data, calculate the product of the first sub-data and its corresponding parameter value to obtain the target sub-data corresponding to the first sub-data; The weighted average of each third sub-data in the first state information and the corresponding target sub-data in the multiple target sub-data is obtained to obtain the multiple target state information.

8. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the user behavior recognition method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the user behavior recognition method as described in any one of claims 1-6.

10. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the user behavior recognition method as described in any one of claims 1-6.

Images