User value information identifying and pushing method based on multiple dimensions

Abstract

The embodiment of the invention discloses a user value information identifying and pushing method based on multiple dimensions. A specific embodiment of the method comprises the steps that a user value model processing end obtains a user association factor information sample set, a value transfer factor information sample set, a value type factor information sample set and a user portrait factor information sample set according to a user association factor information sample set, a value transfer factor information sample set and a user portrait factor information sample set; respectively performing model training on the initial user association information state recognition model, the initial value circulation information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model; and the user information pushing end inputs to-be-identified user information into the trained user value information identification model, and sends a user value information identification result to a corresponding user terminal. According to the embodiment, the accuracy and effectiveness of user value information identification are improved, re-pushing is not needed, waste of communication resources is reduced, the accuracy of abnormal user information identification is improved, and accurate interception of abnormal information is realized.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology, and in particular to a method for identifying and pushing user value information based on multiple dimensions. Background Art

[0002] Currently, user value information identification can be a technology that identifies massive amounts of user information and analyzes and identifies users, and can perform user management based on the results of user value information identification. However, there is a large amount of abnormal user information in the massive amount of user information, and the processing of a large amount of abnormal user information is an important factor affecting user information security and the traffic interception load of each terminal. At present, when using a multi-dimensional user value information identification and push method, the method that can be used is: manually preset a set of user information analysis rules based on experience, determine the user value information of the user information through the user information analysis rules, and then push the identified user value information to each client.

[0003] However, the above method often encounters the following technical problems: due to the subjectivity and limitations of user information analysis rules, it is impossible to effectively adapt to the real-time changes of user information, and in the face of massive user information, it is impossible to effectively capture all relevant information, which may cause important information to be ignored. There is a large amount of redundant and erroneous user information, resulting in errors in the pushed information, and the need to push the information again, causing a serious waste of system communication resources; in addition, only abnormal alarms are issued for abnormal user information, and the behavior of abnormal user information cannot be intercepted and locked, resulting in a large amount of abnormal traffic, resulting in a high load on each terminal, reducing the security of user information and the stability of each terminal.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0005] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0006] Some embodiments of the present disclosure propose a multi-dimensional method for identifying and pushing user value information to solve one or more of the technical problems mentioned in the above background technology section.

[0007] In a first aspect, some embodiments of the present disclosure provide a method for identifying and pushing user value information based on multiple dimensions, the method comprising: a user value model processing end constructing an initial user value information identification model, wherein the initial user value information identification model comprises: an initial user association information state identification model, an initial value flow information state identification model, an initial value type information state identification model, and an initial user portrait state identification model; a user indicator data processing end acquiring an initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set; the user indicator data processing end processes the initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set; The user-related information sample set, the initial value flow information sample set, the initial value type information sample set and the initial user portrait information sample set are respectively subjected to data factor extraction processing to generate an initial user-related factor information sample set, an initial value flow factor information sample set, an initial value type factor information sample set and an initial user portrait factor information sample set; the user indicator data processing end performs factor selection processing on the initial user-related factor information sample set, the initial value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set to obtain the selected initial user-related factor information sample set, the initial value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set. The value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set are used as the user association factor information sample set, the value flow factor information sample set, the value type factor information sample set and the user portrait factor information sample set; the user value model processing end performs model training on the initial user association information state recognition model, the initial value flow information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model according to the user association factor information sample set, the value flow factor information sample set, the value type factor information sample set and the user portrait factor information sample set, respectively, to obtain the training completion The user-related information status recognition model, the value circulation information status recognition model, the value type information status recognition model and the user portrait status recognition model are provided; the user information push end inputs the user information to be recognized into the trained user value information recognition model to obtain the user value information recognition result, and in response to determining that the user value information recognition result meets the information sending condition, sends the user value information recognition result to the corresponding user terminal, wherein the trained user value information recognition model includes: the user-related information status recognition model, the value circulation information status recognition model, the value type information status recognition model and the user portrait status recognition model.

[0008] In a second aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0009] In a third aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method described in any implementation manner in the first aspect is implemented.

[0010] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the multi-dimensional user value information identification and push method of some embodiments of the present disclosure, the accuracy and effectiveness of user value information identification are improved, no re-push is required, the waste of communication resources is reduced, and the accuracy of identifying abnormal user information is improved, and the accurate interception of abnormal user information is achieved, the security of user information is improved, and the load of each terminal is reduced. Specifically, the reason for the serious waste of system communication resources is that: due to the subjectivity and limitations of the user information analysis rules, it is impossible to effectively adapt to the real-time changes of user information, and in the face of massive user information, it is impossible to effectively capture all relevant information, which may cause important information to be ignored, and there is a large amount of redundant and erroneous user information, resulting in errors in the pushed information, and the need to re-push the information, resulting in a serious waste of system communication resources; in addition, only abnormal alarms are issued for abnormal user information, and the behavior of abnormal user information cannot be intercepted and locked, resulting in a large amount of abnormal traffic, resulting in a high load on each terminal, reducing the security of user information and the stability of each terminal. Based on this, in some embodiments of the present disclosure, the user value information identification and push method based on multi-dimensionality, the user value model processing end constructs an initial user value information identification model, wherein the above-mentioned initial user value information identification model includes: an initial user association information state identification model, an initial value flow information state identification model, an initial value type information state identification model, and an initial user portrait state identification model. Here, the constructed initial user value information identification model facilitates the subsequent data processing of the user information to be identified based on big data technology. The user indicator data processing end obtains the initial user association information sample set, the initial value flow information sample set, the initial value type information sample set, and the initial user portrait information sample set. Here, the obtained initial user association information sample set, initial value flow information sample set, initial value type information sample set, and initial user portrait information sample set can be used for subsequent data factor extraction processing. The user indicator data processing end performs data factor extraction processing on the initial user association information sample set, the initial value flow information sample set, the initial value type information sample set and the initial user portrait information sample set, respectively, to generate an initial user association factor information sample set, an initial value flow factor information sample set, an initial value type factor information sample set and an initial user portrait factor information sample set. Here, the data factor extraction processing can remove redundant error information in each sample set, improve the quality of each sample set, and each sample obtained can represent different aspects of user information from multiple dimensions, so as to facilitate the subsequent more comprehensive capture of user information features.The user indicator data processing end performs factor selection processing on the initial user association factor information sample set, the initial value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set, respectively, to obtain the selected initial user association factor information sample set, the initial value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set as the user association factor information sample set, the value flow factor information sample set, the value type factor information sample set and the user portrait factor information sample set. Here, the factor selection processing can effectively identify the potential characteristics and correlations between each sample set, so as to facilitate the subsequent determination of the proportion of each sample set in the identification of user value information. The user value model processing end performs model training on the initial user association information state recognition model, the initial value flow information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model according to the user association factor information sample set, the value flow factor information sample set, the value type factor information sample set and the user portrait factor information sample set, respectively, to obtain the trained user association information state recognition model, the value flow information state recognition model, the value type information state recognition model and the user portrait state recognition model. Here, model training can make the various models included in the user value information recognition model more suitable for the current scenario, improve the accuracy of user value information recognition in the current scenario, and the model training can be automatically updated and optimized, which can adapt to the problem of real-time changes in user information and improve the timeliness and effectiveness of user value information recognition. The user information push terminal inputs the user information to be identified into the trained user value information identification model to obtain the user value information identification result, and in response to determining that the user value information identification result meets the information sending condition, sends the user value information identification result to the corresponding user terminal, wherein the trained user value information identification model includes: user association information state identification model, value flow information state identification model, value type information state identification model and user portrait state identification model. Here, the trained user value information identification model includes: user association information state identification model, value flow information state identification model, value type information state identification model and user portrait state identification model to identify user value information, and perform more comprehensive user feature identification from different dimensions. The integration of different feature factors can improve the accuracy of identification, improve the accuracy of information push results, and reduce the waste of transmission resources for information push. Therefore, the construction of the model avoids the subjective bias that may exist when manually setting rules, and ensures the objectivity and consistency of user value information identification. At the same time, it can automatically update and optimize in real time according to the user information to be identified through the model, which improves the relevance and effectiveness of the user value information identification results. The training of the model reduces the situation where the business frequently manually intervenes in the rules.The model is used to adjust and optimize the risk weights and integrate various sample sets. This integration encompasses the mutual influence between different sample sets and is more flexible and targeted. The customer portrait model is integrated to cover more dimensional customer features, making the identification of user value information more comprehensive and robust. In addition, by using big data technology and machine learning algorithms, detailed customer features are mined from multiple dimensions to provide more comprehensive and in-depth support information for the identification of personalized customer value information. At the same time, by covering multi-dimensional features, the model can more robustly identify customer information and maintain high accuracy and stability as much as possible when user information is constantly changing. In addition, the accuracy of pushing user value information identification results is improved, and there is no need to re-push, which reduces the waste of communication resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0012] Figure 1 is a flowchart of some embodiments of the method for identifying and pushing user value information based on multiple dimensions according to the present disclosure;

[0013] Figure 2 It is a schematic diagram of the collinearity results of credit information and transaction information in the multi-dimensional user value information identification and push method disclosed in the present invention;

[0014] Figure 3 It is a flow chart of an interpolation transformation based on the initial user portrait state recognition result in the multi-dimensional user value information recognition and push method disclosed in the present invention;

[0015] Figure 4 It is a schematic diagram of the score distribution of a credit risk and a transaction risk in the method for identifying and pushing user value information based on multiple dimensions disclosed in the present invention;

[0016] Figure 5 It is a schematic diagram of the confusion results of the model under different groups in the multi-dimensional user value information identification and push method disclosed in the present invention;

[0017] Figure 6 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0019] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0020] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0023] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0024] Figure 1 The process 100 of some embodiments of the method for identifying and pushing user value information based on multiple dimensions according to the present disclosure is shown. The method for identifying and pushing user value information based on multiple dimensions includes the following steps:

[0025] Step 101: The user value model processing end constructs an initial user value information recognition model.

[0026] In some embodiments, the user value model processing end constructs an initial user value information recognition model. The initial user value information recognition model includes: an initial user association information state recognition model, an initial value flow information state recognition model, an initial value type information state recognition model, and an initial user portrait state recognition model. The initial user association information state recognition model may be a pre-established untrained user association information state recognition model. The user association information state recognition model may be a neural network model that takes user association information as input and outputs a user association information state recognition result. The user association information state recognition result may represent a score of the user association information. For example, the initial user association information state recognition model may be an untrained Z score formula, a ZETA scoring model, a Credit Metrics model, a Credit Risk+ model, or a Credit Portfolio View model. For example, user association information may refer to information related to user credit. The initial value flow information state recognition model may be a pre-established untrained value flow information state recognition model. The value flow information state recognition model may be a neural network model that takes value flow information as input and outputs a value flow information state recognition result. The result of the value flow information status identification can represent the score of the value flow information. For example, the initial value flow information status identification model can be an untrained KMV model, JP Morgan's VAR model (Value At Risk), RAROC model (Risk-Adjusted Return on Capital), and EVA model (Economic Value Added). For example, value flow information can refer to value transaction risk information. The initial value type information status identification model can be a pre-established untrained value type information status identification model. The value type information status identification model can be a neural network model that takes value type information as input and takes value type information status identification result as output. The value type information status identification result can represent the score of value type information. For example, the initial value type information status identification model can be an untrained scorecard model or a fraud detection model. The above-mentioned fraud detection model can be a model that first uses a label propagation algorithm to determine the potential risk value type information in the value type graph corresponding to the value type information; then, inputs the potential risk value type information into the MVGNN model (Multi-View Graph Neural Network) to obtain a value type information status recognition result.For example, value type information may refer to value variety information (macroeconomic conditions, market economic conditions, industry economic conditions, monetary policy, fiscal policy, stock, bond and exchange market data). The initial user portrait state recognition model may be a pre-established, untrained user portrait state recognition model. The user portrait state recognition model may be a neural network model that takes the user portrait as input and the user portrait state recognition result as output. The user portrait state recognition result may represent the score of the user portrait. For example, the initial user portrait state recognition model may be an untrained RFM model (customer relationship management model) or AARRR model (pirate model). For example, the user portrait may refer to the basic information of the user. The user value model processing end may be a computing terminal for constructing a user value information recognition model.

[0027] In step 102, the user indicator data processing end obtains an initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set.

[0028] In some embodiments, the user indicator data processing end may obtain an initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set from a pre-set database. Among them, the user indicator data processing end may be a terminal for processing an initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set. The database may be a database that pre-stores each user association information sample, value flow information sample, value type information sample, and user portrait information sample. The initial user association information sample may be a sample for identifying user credit risk. The initial user association information sample may include, but is not limited to, at least one of the following: the industry to which the user's company belongs, the registered scale of the enterprise, (abnormal) operating conditions, transaction data, account opening date, transaction win rate in the past year, turnover rate, investment type, and risk tolerance. The initial value flow information sample may be a sample of information on the circulation of virtual value items (financial products). The initial value circulation information sample may include but is not limited to at least one of the following: holding information of virtual value items (position structure information), category information of virtual value items, credit risk level of the company where the virtual value items are located, and circulation volume (transaction volume) of virtual value items. The initial value type information sample may be a sample of type information of virtual value items. The initial value type information sample may include but is not limited to at least one of the following: industry economic conditions, monetary policy, fiscal policy, stock, bond and foreign exchange market data, industry supply and demand. The user portrait information sample may represent a sample of user behavior information. The user portrait information sample may include but is not limited to at least one of the following: user basic information data, income information, expenditure information, transaction data, etc.

[0029] Step 103, the user indicator data processing end performs data factor extraction processing on the initial user association information sample set, the initial value flow information sample set, the initial value type information sample set and the initial user portrait information sample set, respectively, to generate an initial user association factor information sample set, an initial value flow factor information sample set, an initial value type factor information sample set and an initial user portrait factor information sample set.

[0030] In some embodiments, the user indicator data processing end performs data factor extraction processing on the initial user association information sample set, the initial value flow information sample set, the initial value type information sample set and the initial user portrait information sample set, respectively, to generate an initial user association factor information sample set, an initial value flow factor information sample set, an initial value type factor information sample set and an initial user portrait factor information sample set. Among them, the initial user association factor information sample set may be a sample set obtained after removing redundant and erroneous data from the initial user association information sample set. The initial value flow factor information sample set may be a sample set obtained after removing redundant and erroneous data from the initial value flow information sample set. The initial value type factor information sample set may be a sample set obtained after removing redundant and erroneous data from the initial value type information sample set. The initial user portrait factor information sample set may be a sample set obtained after removing redundant and erroneous data from the initial user portrait information sample set.

[0031] For example, the data factor extraction process may refer to: first, determining the sample distribution information of the above-mentioned initial user association information sample set, the above-mentioned initial value flow information sample set, the above-mentioned initial value type information sample set and the above-mentioned initial user portrait information sample set. Then, designing a statistical algorithm (or a regular algorithm) to identify and preprocess the above-mentioned initial user association information sample set, the above-mentioned initial value flow information sample set, the above-mentioned initial value type information sample set and the above-mentioned initial user portrait information sample set one by one, and perform validity verification to meet the model input and analysis requirements. Among them, the above-mentioned preprocessing may include but is not limited to at least one of the following: data outlier (abnormal character) processing, data deduplication, missing value processing (Nan value (Not a Number, non-number), inf value (infinite, infinity), None, empty character), extreme value processing, data standardization and data conversion. Extreme value processing can be to remove the standard deviation that deviates from the mean value by more than 3 times. If there is unstructured data in the above-mentioned initial user association information sample set, the above-mentioned initial value flow information sample set, the above-mentioned initial value type information sample set and the above-mentioned initial user portrait information sample set, first use the NLP algorithm (Natural Language Processing, natural language processing algorithm) to convert the unstructured data into structured data and then perform identification and preprocessing.

[0032] Step 104, the user indicator data processing end performs factor selection processing on the initial user-associated factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set, respectively, to obtain the selected initial user-associated factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set as the user-associated factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set.

[0033] In some embodiments, the above-mentioned user indicator data processing end performs factor selection processing on the initial user-associated factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set, respectively, to obtain the selected initial user-associated factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set as the user-associated factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set.

[0034] For example, the initial user-related factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set can be processed by discretization method based on chi-square test / discretization method based on decision tree / significance measurement method based on information entropy and information entropy-like indicators, respectively, to obtain the selected initial user-related factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set as the user-related factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set. That is, the standard and method of factor selection can be set according to actual needs.

[0035] Optionally, the user indicator data processing end sends the user association factor information sample set, value circulation factor information sample set, value type factor information sample set and user portrait factor information sample set to the user value model processing end.

[0036] Step 105: The user value model processing end performs model training on the initial user association information state recognition model, the initial value flow information state recognition model, the initial value type information state recognition model, and the initial user portrait state recognition model according to the user association factor information sample set, the value flow information sample set, the value type factor information sample set, and the user portrait factor information sample set, respectively, to obtain trained user association information state recognition models, value flow information state recognition models, value type information state recognition models, and user portrait state recognition models.

[0037] In some embodiments, the user value model processing end performs model training on the initial user association information state recognition model, the initial value flow information state recognition model, the initial value type information state recognition model, and the initial user portrait state recognition model according to the user association factor information sample set, the value flow information sample set, the value type factor information sample set, and the user portrait factor information sample set, respectively, to obtain the trained user association information state recognition model, the value flow information state recognition model, the value type information state recognition model, and the user portrait state recognition model.

[0038] In the process of adopting technical solutions to solve the above-mentioned technical problem 1, the following technical problem 2 is often accompanied: how to balance the weights of the output results of different recognition models to improve the accuracy of user value information recognition based on the user value information recognition model, intercept and lock abnormal user information, improve the stability and security of each terminal, and improve the security of user information. For the above-mentioned technical problem 2, the conventional solution is generally: determine the weights of the output results of different recognition models through an objective assignment method to obtain the user value information recognition result. Then, determine the abnormal user information set through the user value information recognition result, and perform alarm processing on the abnormal user information set. However, the above conventional solutions still have the following problems: Since the objective assignment method usually over-relies on the surface features and statistical laws of the sample set, and only performs simple processing on different sample sets, there are a large number of redundant erroneous data in the sample set, and it is impossible to identify the deep features within the sample set and the correlation between the sample sets, and it is unable to adapt to real-time changes and updates, resulting in low accuracy of the determined weights, low accuracy of user value information identification results, low accuracy of user abnormal identification, and a large amount of abnormal traffic on the operation of each terminal, resulting in a high load on each terminal, reducing the stability and security of each terminal, user information leakage, and low security of user information. The inventors took into account the shortcomings of conventional solutions and combined with the advantages / technical status of the user value information identification technology owned by the inventor's company, we decided to adopt the following solution:

[0039] In practice, the user value model processing end can respectively train the initial user association information state recognition model, the initial value flow information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model through the following steps:

[0040] The first step is to select a target user-related factor information sample from the above user-related factor information sample set. For example, a user-related factor information sample can be randomly selected from the above user-related factor information sample set as the target user-related factor information sample.

[0041] The second step is to input the target user association factor information sample into the initial user association information state recognition model to obtain the initial user association information state recognition result. The initial user association information state recognition result represents the state score of the user association factor information sample. The higher the state score, the lower the risk represented by the user association factor information.

[0042] The third step is to perform interpolation transformation on the initial user association information state recognition result to obtain the initial user association information transformation interpolation.

[0043] For example, the above initial user association information state recognition result can be interpolated and transformed in the following manner:

[0044] First, the collinearity test is performed on the credit information and transaction information in the user association factor information sample and the value flow factor information sample to obtain the collinearity test results. It should be noted that the collinearity test can grasp the degree of correlation between credit information and transaction information, thereby ensuring that there is no overlapping redundant data in the credit information and transaction information, and determining the diversity of the sample to improve the accuracy of subsequent interpolation. Constructing the customer margin model from the dimensions of credit information and transaction information will not cause multicollinearity problems. For example, for a sample in a certain scenario, the collinearity results of credit information and transaction information are as follows: Figure 2 The example shown. Figure 2 In the figure, the horizontal axis can be credit information, and the vertical axis can be transaction information. corr can represent the correlation coefficient (Pearson correlation coefficient) between credit information and transaction information. Vif can represent the degree of multicollinearity between credit information and transaction information. Figure 2 In the figure, corr = 0.08, vif = 1.93, which shows that there is no obvious collinearity between the two.

[0045] Then, in response to determining that the collinearity test result is greater than or equal to a preset collinearity test threshold, a model target value estimation result is obtained through interval interpolation function approximation.

[0046] Among them, the preset collinearity test threshold can be a pre-set critical value for determining whether two samples are multicollinear. The above interval interpolation function can represent a function for interpolating the initial user association information state recognition result. By statistically analyzing the risk score, interpolation calculation is performed based on the credit risk score to obtain the interpolation result corresponding to the credit risk. Interpolation calculation is performed based on the transaction risk score to obtain the interpolation result corresponding to the transaction risk. Interpolation transformation is performed in the following manner:

[0047] Assume that the risk sample sequence (user association factor information sample and value flow factor information sample) is: X = {x 1 , x 2 , ..., x n}, where x i is the risk score of the risk sample (i.e., the initial user association information status identification result), and R i ≤x i ≤R m , for each risk x in X, y can be calculated by the following interpolation function:

[0048]

[0049] Among them, the sequence m c ={y 1 ,y 2 , ..., y n} represents the initial user association information transformation interpolation sequence after interpolation. Wherein, c is a constant, and its value is c=0; [M i , M m ] is the business risk margin range corresponding to the preset risk sample sequence. m It is the maximum value of the business risk margin. i is the minimum value of the business risk margin. i Represents the minimum value in the risk sample sequence. R m Represents the maximum value in the risk sample sequence.

[0050] The fourth step is to determine the correlation difference between the initial user association information transformation interpolation and the corresponding user association factor information sample label. For example, the correlation difference between the initial user association information transformation interpolation and the corresponding user association factor information sample label can be determined by a cross entropy loss function, a hinge loss function, or a cosine loss function. The user association factor information sample label can represent a preset standard transformation interpolation.

[0051] In step 5, in response to determining that the association difference value is less than or equal to a preset difference value, the initial user association information state recognition model is determined as the trained user association information state recognition model. The preset difference value may be a pre-set critical value for determining whether an association difference is formed.

[0052] Step 6: Select a target value circulation factor information sample from the above value circulation factor information sample set. For example, a value circulation factor information sample can be randomly selected from the above value circulation factor information sample set as the target value circulation factor information sample.

[0053] In the seventh step, the target value flow factor information sample is input into the initial value flow information state identification model to obtain the initial value flow information state identification result, wherein the initial value flow information state identification result represents the state score of the value flow factor information sample.

[0054] The eighth step is to perform interpolation transformation on the above initial value flow information state recognition result to obtain the initial value flow information transformation interpolation.

[0055] For example, the above initial value flow information state recognition result can be interpolated and transformed in the following way:

[0056] First, the collinearity test is performed on the credit and transaction of the user association factor information sample and the value flow factor information sample. Ensure that the customer margin model is constructed from the customer credit and transaction dimensions without multicollinearity problems. For example, for a sample in a certain scenario, the collinearity results of credit and transaction are as follows: Figure 2 For example, customer credit risk and transaction risk, corr = 0.08, vif = 1.93, indicating that there is no obvious collinearity between the two.

[0057] Secondly, in this scheme, the model target value estimation result is obtained through interval interpolation function approximation.

[0058] Among them, by statistically analyzing the risk score, interpolation calculation is performed based on the credit risk score to obtain the interpolation result corresponding to the credit risk. Interpolation calculation is performed based on the transaction risk score to obtain the interpolation result corresponding to the transaction risk. This is done in the following way:

[0059] Assume that the risk sample sequence (user association factor information sample and value flow factor information sample) is: X = {x 1 , x 2 , ..., x n}, where x i is the risk score of the risk sample (i.e., the initial user association information status identification result), and R i ≤x i≤R m , for each risk x in X, y can be calculated by the following interpolation function:

[0060]

[0061] Among them, the sequence m c ={y 1 ,y 2 , ..., y n} represents the initial user association information transformation interpolation sequence after interpolation. Wherein, c is a constant, and in general scenarios, c=0; [M i , M m ] is the business risk margin range corresponding to the preset risk sample sequence. i Represents the minimum value in the risk sample sequence. m Represents the maximum value in the risk sample sequence.

[0062] The ninth step is to determine the value difference between the above-mentioned initial value flow information transformation interpolation and the corresponding value flow factor information sample label. For example, the value difference between the above-mentioned initial value flow information transformation interpolation and the corresponding value flow factor information sample label can be determined by a cross entropy loss function, a hinge loss function, or a cosine loss function.

[0063] In step 10, in response to determining that the value difference value is less than or equal to the preset value difference value, the initial value flow information state recognition model is determined as the trained value flow information state recognition model. The preset value difference value may be a pre-set critical value for determining whether a value difference is formed.

[0064] Step 11: Select a target value type factor information sample from the above value type factor information sample set. For example, a value type factor information sample can be randomly selected from the above value type factor information sample set as the target value type factor information sample.

[0065] Step 12: Input the target value type factor information sample into the initial value type information state recognition model to obtain the initial value type information state recognition result, wherein the initial value type information state recognition result represents the state score of the value type factor information sample.

[0066] In the thirteenth step, interpolation transformation is performed on the above-mentioned initial value type information state recognition result to obtain the initial value type information transformation interpolation.

[0067] Step 14: Determine the value type difference between the initial value type information transformation interpolation and the corresponding value type factor information sample label. For example, the value type difference between the initial value type information transformation interpolation and the corresponding value type factor information sample label can be determined by a cross entropy loss function, a hinge loss function, or a cosine loss function.

[0068] In step 15, in response to determining that the value type difference value is less than or equal to the preset value type difference value, the initial value type information state recognition model is determined as the trained value type information state recognition model. The preset value type difference value may be a pre-set critical value for determining whether a value type difference is formed.

[0069] Step 16: Select a target user portrait factor information sample from the above user portrait factor information sample set. A user portrait factor information sample can be randomly selected from the above user portrait factor information sample set as the target user portrait factor information sample.

[0070] Step 17: Input the target user portrait factor information sample into the initial user portrait state recognition model to obtain the initial user portrait state recognition result, wherein the initial user portrait state recognition result represents the state score of the user portrait factor information sample.

[0071] In the eighteenth step, interpolation transformation is performed on the above initial user portrait state recognition result to obtain the initial user portrait transformation interpolation.

[0072] For example, Figure 3 In the example, the value type (variety) risk can be divided into several levels (such as 1 to 6), and the different levels can be merged and regrouped (discretized and binned). The grouping method is determined by IV (information value) calculation, and then the Woe (weight of evidence) under each group is calculated. The Woe interpolation transformation is used to obtain the margin interpolation result corresponding to the variety risk. The interpolation method can be linear interpolation.

[0073] Through a small number of known samples, using the linear regression algorithm, w_p, w_c, w_t, and w_f are calculated and determined. Among them, w_p can represent the weight value of the initial user portrait transformation interpolation, w_c can represent the weight value of the initial user association information transformation interpolation, w_t can represent the weight value of the initial value flow information transformation interpolation, and w_f can represent the weight value of the initial value type information transformation interpolation. w_p, w_c, w_t, and w_f can also be set or adjusted according to the business focus. The user value information identification result (margin_poi value, the middle value of the margin) is obtained. The distribution of margin_poi and customer credit risk and transaction risk scores is shown in the following figure. Figure 4 As shown in the example, margin_poi = w_p*s_p+w_c*s_c+w_t*s_t+w_f*s_f, s_p can represent the initial user portrait transformation interpolation. s_c can represent the initial user association information transformation interpolation. s_t can represent the initial value flow information transformation interpolation. s_f can represent the initial value type information transformation interpolation.

[0074] In step 19, the user portrait difference value between the initial user portrait transformation interpolation and the corresponding user portrait factor information sample label is determined. For example, the user portrait difference value between the initial user portrait transformation interpolation and the corresponding user portrait factor information sample label can be determined by a cross entropy loss function, a hinge loss function, or a cosine loss function.

[0075] In step 20, in response to determining that the user portrait difference value is less than or equal to a preset user portrait difference value, the initial user portrait state recognition model is determined as a trained user portrait state recognition model. The preset user portrait difference value may be a pre-set critical value for determining whether a user portrait difference is formed.

[0076] It should be noted that the user value model processing end can use the above steps 1 to 20 to calculate the margin_poi of all samples, and then establish a user value information recognition model (personalized margin evaluation adaptation model) for the user:

[0077] M(s_p,s_c,s_t,s_f))=μ / [1+e- (w_p*s_p+w_c*s_c+w_t*s_t+w__f*s_f+inte)α+β ],

[0078] In order to ensure that the model results are white-box and the user value information identification results can be compared and traced, the model M(·) can be a user value information identification model. The logistic regression model framework is used for construction and training to obtain the values ​​of w_p, w_c, w_t, w_f, and inte. Among them, μ, α, and β are adjustment parameters, which can be flexibly set and adjusted by business personnel according to business development needs in different business scenarios. The model also needs to be processed as follows. Inte can be understood as the intercept used to correct the model result curve.

[0079] Model parameter interpretation performance: In order to ensure the effectiveness of parameter estimation and significance testing, and to ensure the predictive performance of the model, the model uses HC0 robust covariance estimation or RANSAC consistency method to deal with the heteroscedasticity of the model. First, RANSAC is used to identify normal values ​​and outliers in the data set to obtain a relatively clean data subset. Linear regression is applied to the relatively "pure" data subset, and HC0 is used to calculate the robust standard error to reduce the impact of outliers on model parameter estimation and ensure that statistical tests (such as t-test, F-test, etc.) are effective.

[0080] Model estimation performance: Since the model output is an integer or a discrete margin value of 0.5, the target value and the predicted value are divided into intervals according to the model target numerical distribution, and the model performance is evaluated according to the rank correlation evaluation method under different level intervals.

[0081] Some statistical results of the model are as follows:

[0082] Model error statistics

[0083] Mae Rmse Corr Corrúic 0.37 0.42 0.87 0.90

[0084] Among them, Mae (Mean Absolute Error) can represent the average value of the absolute error between the predicted value and the true value output by the user value information identification model. Rmse (Root Mean Squared Error) can be the square root of the mean square error (the average value of the sum of the squares of the differences between the predicted value and the true value). Corr (Correlation Coefficient) can be a correlation coefficient, that is, the Pearson correlation coefficient. Corr_ic (Correlation of information coefficient) can measure the correlation index between the predicted value and the true value, that is, it can characterize the effectiveness of the predicted value output by the user value identification model.

[0085] The confusion results of the model under different groups are as follows Figure 5 The example shown.

[0086] In the 21st step, the user information to be identified is input into the trained user value information identification model, which includes: user association information state identification model, value flow information state identification model, value type information state identification model and user portrait state identification model to obtain the user value information identification result.

[0087] In step 22, in response to detecting that the user value information identification result pushed by the user value model processing end meets the user value information alarm condition, the user information push end marks the user account corresponding to the user information to be identified as abnormal, and sends an alarm message to the associated user information abnormality processing terminal.

[0088] In the twenty-third step, the user information exception processing terminal responds to the alarm information sent by the user information push terminal by locking the electronic signature and digital authentication tool corresponding to the alarm information, and locking the user account corresponding to the alarm information.

[0089] The technical contents of the first step to the twenty-third step mentioned above, as an inventive point of an embodiment of the present disclosure, solve the second technical problem: "Since the objective assignment method usually relies too much on the surface features and statistical laws of the sample set, and only performs simple processing on different sample sets, there are a large number of redundant and erroneous data in the sample set, and it is impossible to identify the deep features within the sample set and the correlation between the sample sets, and it is impossible to adapt to real-time changes and updates, resulting in low accuracy in the determined weights, low accuracy in the user value information identification results, low accuracy in user abnormal identification, and a large amount of abnormal traffic on the operation behavior of each terminal, which leads to a high load on each terminal, reduces the stability and security of each terminal, and leaks user information, and the security of user information is low." The factors that lead to the increase of the load of each terminal, the reduction of the stability and security of each terminal, the leakage of user information, and the low security of user information are often as follows: Since the objective assignment method usually over-relies on the surface characteristics and statistical laws of the sample set, and only performs simple processing on different sample sets, there are a large number of redundant error data in the sample set, and the deep characteristics inside the sample set and the association between the sample sets cannot be identified, and it cannot adapt to the situation of real-time changes and updates, resulting in low accuracy of the determined weight, low accuracy of the user value information recognition result, low accuracy of user abnormal recognition, and a large number of abnormal traffic operations on each terminal. If the above factors are solved, it can achieve the effect of reducing the load of each terminal, improving the stability and security of each terminal, reducing the leakage of user information to a certain extent, and improving the security of user information. In order to achieve this effect, the present disclosure firstly, by training the model of interpolation transformation of the user association information state recognition model, the value flow information state recognition model, the value type information state recognition model and the user portrait state recognition model, the user value information recognition result of the customer in the current state can be generated in real time. The real-time performance is reflected in the fact that the system can instantly calculate the standard of the customer's user value information (margin) based on the latest user association information sample set, value flow information sample set, value type information sample set and user portrait information sample set, and can understand the customer's latest risk status and make decisions at the first time. At the same time, as the user information changes in real time, the system can automatically update to ensure that it always reflects the latest situation. This dynamic update mechanism avoids the lag problem that may be caused by fixed rules in traditional methods. In addition, it can run continuously and automatically, automatically learn to adjust the risk contribution of various risks, and improve the efficiency of margin review and margin determination. The system can run continuously and automatically without human intervention, reducing errors and delays.Then, through the user value information identification results identified by the user association information status identification model, the value flow information status identification model, the value type information status identification model and the user portrait status identification model, the user account and the user value operation terminal that meet the user value information alarm conditions in the user value information identification results are locked, and the terminal corresponding to the abnormal user value information identification results can be quickly tracked, which not only avoids the abnormal user value operation terminal from sending invalid information, but also reduces the number of invalid information processed by the system, thereby reducing the waste of computing power resources of each terminal, reducing the locking processing of abnormal traffic by each terminal, reducing the load of each terminal and improving the stability of system operation. Finally, because the user account and the user value operation terminal can be locked, the loss of value deposit is avoided and the security of user information is improved.

[0090] Step 106, the user information push end inputs the user information to be identified into the trained user value information identification model to obtain the user value information identification result, and in response to determining that the user value information identification result meets the information sending condition, sends the user value information identification result to the corresponding user terminal.

[0091] In some embodiments, the user information push end inputs the user information to be identified into the trained user value information identification model to obtain the user value information identification result, and in response to determining that the user value information identification result satisfies the information sending condition, sends the user value information identification result to the corresponding user terminal. Among them, the trained user value information identification model includes: user association information state identification model, value flow information state identification model, value type information state identification model and user portrait state identification model. The user value information identification result may refer to the value guarantee information (for example, deposit) corresponding to the user information to be identified output by the user value information identification model. The information sending condition may refer to the value guarantee information represented by the user value information identification result being within a preset range.

[0092] Optionally, in response to determining that the user value information identification result meets the user value information alarm condition, the user information push end marks the user account corresponding to the user information to be identified as abnormal, and sends an alarm message to the associated user information abnormality processing terminal.

[0093] In some embodiments, in response to determining that the user value information identification result satisfies the user value information alarm condition, the user information push terminal marks the user account corresponding to the user information to be identified as abnormal, and sends an alarm message to the associated user information abnormality processing terminal. The user value information alarm condition may refer to: the value represented by the user value information identification result is not within a preset range. The user information abnormality processing terminal may refer to an account supervision terminal that is communicatively connected to the user information push terminal. The alarm message may indicate that the user account is abnormal.

[0094] Optionally, in response to receiving the alarm information, the user information exception processing terminal locks the electronic signature and digital authentication tool end corresponding to the alarm information, and locks the user account corresponding to the alarm information.

[0095] In some embodiments, the user information exception processing terminal, in response to receiving the alarm information, locks the user value operation terminal corresponding to the alarm information, and locks the user account corresponding to the alarm information. The electronic signature and digital authentication tool terminal can represent the value processing terminal corresponding to the user account. For example, the electronic signature and digital authentication tool terminal can represent the U shield corresponding to the user account. The locking process can represent a function locking process, that is, the user value operation terminal can no longer perform value information operations (for example, amount transfer).

[0096] In practice, the user information push terminal can input the user information to be identified into the trained user value information identification model through the following steps:

[0097] The first step is to input the user information to be identified into the user-related information state identification model included in the user value information identification model to obtain the user-related information state identification result.

[0098] The second step is to input the user information to be identified into the value flow information status identification model included in the user value information identification model to obtain the value flow information status identification result.

[0099] The third step is to input the user information to be identified into the value type information state identification model included in the user value information identification model to obtain the value type information state identification result.

[0100] The fourth step is to input the above-mentioned user information to be identified into the user portrait status identification model included in the above-mentioned user value information identification model to obtain the user portrait status identification result.

[0101] The fifth step is to generate a user value information recognition result based on the above user-related information status recognition result, the above value flow information status recognition result, the above value type information status recognition result and the above user portrait status recognition result.

[0102] The fifth step may include the following sub-steps:

[0103] The first sub-step is to perform interpolation transformation on the user-related information state recognition result to generate a user-related information state value.

[0104] The second sub-step is to perform interpolation transformation on the above-mentioned value flow information status identification result to generate a value flow information status value.

[0105] The third sub-step is to perform interpolation transformation on the above-mentioned value type information state recognition result to generate a value type information state value.

[0106] The fourth sub-step is to perform interpolation transformation on the above user portrait state recognition result to generate a user portrait state value.

[0107] Here, the above-mentioned state values ​​may represent transformation interpolation.

[0108] The fifth sub-step is to generate a user value information recognition result based on the user association information status value, the value flow information status value, the value type information status value and the user portrait status value.

[0109] Here, the implementation method of the fifth sub-step can refer to the step description between the first step to the twentieth step above, which will not be repeated here.

[0110] Among them, the above-mentioned user value model processing end is communicatively connected with the above-mentioned user indicator data processing end, and the above-mentioned user information pushing end is communicatively connected with the above-mentioned user value model processing end and the above-mentioned user indicator data processing end respectively.

[0111] In the process of adopting technical solutions to solve the above technical problem 1, the following technical problem 3 is often accompanied: how to balance the weights of the output results of different recognition models to improve the accuracy of user value information recognition based on the user value information recognition model, intercept and lock abnormal user information, improve the stability and security of each terminal, and improve the security of user information. For the above technical problem 3, the conventional solution is generally: using the particle swarm optimization algorithm to determine the weights of the output results of different recognition models to obtain the user value information recognition results. Then, the abnormal user information set is determined by the user value information recognition results, and the abnormal user information set is alarmed. However, the above conventional solution still has the following problems: due to random initialization in the particle swarm optimization algorithm, it is easy to cause uneven distribution of the initial position of the particle swarm, increase resource waste and prolong computing resources, and the particle swarm optimization algorithm is easy to fall into the local optimal solution, so that the accuracy of the generated distribution weight is low, which leads to a low accuracy of the user value information recognition result, resulting in a low accuracy of user abnormal recognition, and there is a large amount of abnormal traffic on each terminal operation behavior, resulting in a high load on each terminal, reducing the stability and security of each terminal, user information is leaked, and the security of user information is low. The inventors considered the shortcomings of conventional solutions and combined the advantages / technical status of the user value information identification technology owned by the inventor's company. We decided to adopt the following solution:

[0112] In some optional implementations of some embodiments, generating a user value information recognition result according to the user association information status value, the value flow information status value, the value type information status value and the user portrait status value may include the following steps:

[0113] In the first step, a chaos mapping algorithm is used to generate an initialization particle group according to the user-related information status value, the value flow information status value, the value type information status value and the user portrait status value. Each initialization particle in the initialization particle group may include: an initialization speed and an initialization position. The chaos mapping algorithm may be a Circle chaos mapping algorithm. Each initialization particle in the initialization particle group may represent a feasible weight allocation method for the user-related information status value, the value flow information status value, the value type information status value and the user portrait status value. The initialization speed may refer to the direction of movement of the particle from the current position to the next position, that is, the search direction of the particle group in the solution space. The initial position may be the initial weight value corresponding to the user-related information status value, the value flow information status value, the value type information status value and the user portrait status value, that is, a feasible weight allocation scheme.

[0114] The second step is to generate a user value fitness function and a user value constraint function for the user association information status value, the value flow information status value, the value type information status value and the user portrait status value. Among them, the user value fitness function can represent the value of accurately identifying users. For example, the user value fitness function can be the opposite of the recognition accuracy. The smaller the user value fitness function, the higher the user value recognition accuracy. The user value constraint function can be the sum of the individual allocation weights being 1.

[0115] The third step is to perform the following particle swarm determination steps based on the initialized particle swarm:

[0116] Sub-step 1: Determine whether the initialized particle swarm satisfies the above user value constraint function.

[0117] Sub-step 2, in response to determining that the initialized particle group satisfies the above-mentioned user value constraint function, input the initialized particle group into the above-mentioned user value fitness function to obtain an initial user value fitness value set. The initial user value fitness value in the above-mentioned initial user value fitness value set can represent the degree of influence of the distribution weight represented by the particle on the accuracy of user value information recognition. The smaller the above-mentioned initial user value fitness value, the better the fitness of the represented particle.

[0118] Sub-step 3, according to the initial user value fitness value set, determine the initial target position of each initialized particle in the initialized particle swarm and the cluster target position of the initialized particle swarm, and obtain the initial target position set and the cluster target position. Among them, the initial target position in the above-mentioned initial target position set can be the position corresponding to the minimum fitness obtained by inputting the position information obtained by the particle from the initial loop iteration to the current iteration number into the user value fitness function. The above-mentioned cluster target position can be the position corresponding to the minimum user value fitness value in the particle swarm in each loop iteration.

[0119] As an example, the execution subject may first determine the particle position group obtained from the start of the loop iteration to the current loop number for each particle in the particle group to obtain a particle position group set. Secondly, for each particle position group in the particle position group set, determine the minimum value of the fitness value corresponding to the particle position group as the initial target position. Then, determine the user value fitness value set of the particle group at the current iteration number as the current particle user value fitness value set. Finally, determine the current particle user value fitness value with the largest value in the current particle user value fitness value set as the cluster target position.

[0120] Sub-step 4, updating the initialized particle swarm according to the initial target position set and the cluster target position to obtain an updated particle swarm.

[0121] As an example, the execution subject can use the speed update formula to update the speed of each particle in the above-mentioned initialized particle group to obtain an updated speed set. Then, use the position update formula to update the position of each particle in the above-mentioned initialized particle group to obtain an updated position set. Finally, the updated speed set and the updated position set are determined as the updated particle group. Among them, the above-mentioned speed update formula can be expressed as:

[0122]

[0123] in, represents the updated particle velocity after k cycles. 1 Represents the acceleration constant, whose value is 2. Represents the initial target position. represents the position of the i-th j-dimensional vector of the particle. 2 Represents the acceleration constant, whose value is 2. Indicates the cluster target location. rand 1 Indicates a random number in the range [0,1]. 2 Represents a random number in the range [0,1].

[0124] The above position update formula can be expressed as:

[0125]

[0126] Sub-step 5: input the updated particle swarm into the above user value fitness function to obtain an updated user value fitness value set.

[0127] Sub-step 6, based on the initial user value fitness value set and the updated user value fitness value set, the updated particle swarm is screened to obtain a screened cluster position and a screened position set. The screened cluster position may be a position corresponding to the user value fitness value with the smallest user value fitness value in the initial user value fitness value set and the updated user value fitness value set. The screened position in the screened position set may be a position corresponding to the user value fitness value with the smaller value in the initial user value fitness value and the updated user value fitness value.

[0128] As an example, the above-mentioned execution entity may first compare each initial user value fitness value in the above-mentioned initial user value fitness value set with the corresponding updated user value fitness value in the above-mentioned updated user value fitness value set to obtain a comparison result set. Secondly, the comparison result set whose initial user value fitness value is greater than or equal to the updated user value fitness value is filtered out from the above-mentioned comparison result set as the target comparison result set. Thirdly, the initial user value fitness value set corresponding to the above-mentioned target comparison result set and the updated user value fitness value set corresponding to the remaining comparison result set are determined as the target user value fitness value set. Then, the position and speed corresponding to the above-mentioned target user value fitness value set are determined as the filtered position set and the filtered speed set. Finally, the position corresponding to the target user value fitness value with the largest value is filtered out from the above-mentioned target user value fitness value set as the updated cluster position.

[0129] Sub-step 7, determining the number of times the above particle swarm determination step has been executed.

[0130] Sub-step 8, in response to determining that the number of executions exceeds the preset execution threshold, determining the updated particle group as an allocation weight value set, and performing weighted summation on the allocation weight value set, the user association information status value, the value flow information status value, the value type information status value and the user portrait status value to obtain a user value information identification result. The preset execution threshold may be a preset maximum number of executions. For example, the preset execution threshold may be 100.

[0131] In the fourth step, in response to determining that the number of executions does not exceed the preset execution threshold, the screened particle group is determined as the initialization particle group, and the sum of the number of executions and the preset threshold is determined as the number of executions, so as to perform the particle group determination step again. The preset threshold may be a pre-set value. For example, the preset threshold may be 1.

[0132] In the fifth step, in response to detecting that the user value information identification result pushed by the user value model processing end meets the user value information alarm condition, the user information push end marks the user account corresponding to the user information to be identified as abnormal, and sends an alarm message to the associated user information abnormality processing terminal.

[0133] In the sixth step, the user information exception processing terminal locks the electronic signature and digital authentication tool end corresponding to the alarm information in response to receiving the alarm information sent by the user information push terminal, and locks the user account corresponding to the alarm information.

[0134] The above-mentioned first to sixth steps and their related contents, as an inventive point of the embodiment of the present disclosure, solve the technical problem three "Due to the random initialization in the particle swarm optimization algorithm, it is easy to cause the initial position distribution of the particle swarm to be uneven, increase the waste of resources and prolong the computing resources, and the particle swarm optimization algorithm is easy to fall into the local optimal solution, so that the accuracy of the generated allocation weight is low, which in turn leads to the low accuracy of the user value information recognition result, resulting in the low accuracy of the user abnormal recognition, and there are a large number of abnormal traffic operations on each terminal, resulting in the increase of the load of each terminal. The stability and security of each terminal are reduced, and the user information is leaked, and the security of the user information is low". The factors that lead to the increase of the load of each terminal, the reduction of the stability and security of each terminal, the leakage of user information, and the low security of user information are often as follows: Due to the random initialization in the particle swarm optimization algorithm, it is easy to cause the initial position distribution of the particle swarm to be uneven, increase the waste of resources and prolong the computing resources, and the particle swarm optimization algorithm is easy to fall into the local optimal solution, so that the accuracy of the generated allocation weight is low. If the above factors are solved, the load of each terminal can be reduced, the stability and security of each terminal can be improved, the leakage of user information can be reduced to a certain extent, and the security of user information can be improved. In order to achieve this effect, the present invention firstly uses the chaotic mapping algorithm to initialize the particle swarm to obtain the initialized particle swarm, which can obtain evenly distributed particle individuals, reduce the randomness of the particle swarm and the quality of the particle swarm. Secondly, the initial particle swarm is updated through the speed update formula and the position update formula. Since the position update is only related to the initial target position set and the cluster target position, the position update determines the optimal position of the particle as the position after the current iteration update, which can improve the particle swarm's optimization ability, convergence speed and to a certain extent avoid falling into the local optimal solution, reducing the waste of computing resources. Afterwards, the updated particle swarm generates an allocation weight value set, and the above-mentioned allocation weight value set, the above-mentioned user association information status value, the above-mentioned value flow information status value, the above-mentioned value type information status value and the above-mentioned user portrait status value are weighted and summed to obtain the user value information recognition result, which can improve the accuracy of the user value information recognition result. Finally, through the user value information identification results, the existing abnormal user information is alarmed and locked, and the terminal corresponding to the abnormal user value information identification result can be quickly tracked, reducing the amount of invalid information processed by the system and the waste of computing resources of each terminal, reducing the locking processing of abnormal traffic by each terminal, reducing the load of each terminal, improving the stability of system operation, and improving the security of user information.

[0135] Reference below Figure 6 , which shows a structural schematic diagram of an electronic device (eg, an electronic device) 600 suitable for implementing some embodiments of the present disclosure. Figure 6The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0136] like Figure 6 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0137] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 6 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0138] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0139] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0140] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (Hyper Text Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0141] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0142] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a server, a program segment or a part of a code, and the server, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0143] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0144] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A method for identifying and pushing user value information based on multiple dimensions, comprising: The user value model processing end constructs an initial user value information recognition model, wherein the initial user value information recognition model includes: an initial user association information state recognition model, an initial value flow information state recognition model, an initial value type information state recognition model, and an initial user portrait state recognition model; The user indicator data processing end obtains an initial user association information sample set, an initial value flow information sample set, an initial value type information sample set, and an initial user portrait information sample set; The user indicator data processing end performs data factor extraction processing on the initial user association information sample set, the initial value flow information sample set, the initial value type information sample set and the initial user portrait information sample set, respectively, to generate an initial user association factor information sample set, an initial value flow factor information sample set, an initial value type factor information sample set and an initial user portrait factor information sample set; The user indicator data processing end performs factor selection processing on the initial user association factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set, respectively, to obtain the selected initial user association factor information sample set, the initial value circulation factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set as the user association factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set; The user value model processing end performs model training on the initial user association information state recognition model, the initial value circulation information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model according to the user association factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set, respectively, to obtain the trained user association information state recognition model, the value circulation information state recognition model, the value type information state recognition model and the user portrait state recognition model; The user information push end inputs the user information to be identified into the trained user value information identification model to obtain the user value information identification result, and in response to determining that the user value information identification result meets the information sending conditions, sends the user value information identification result to the corresponding user terminal, wherein the trained user value information identification model includes: user-related information status identification model, value flow information status identification model, value type information status identification model and user portrait status identification model.

2. The method according to claim 1, wherein: The method further comprises: In response to determining that the user value information identification result meets the user value information alarm condition, the user information push end marks the user account corresponding to the user information to be identified as abnormal, and sends alarm information to the associated user information abnormality processing terminal; In response to receiving the alarm information, the user information exception processing terminal locks the electronic signature and digital authentication tool end corresponding to the alarm information, and locks the user account corresponding to the alarm information.

3. The method according to claim 1, wherein: The user value model processing end is in communication connection with the user indicator data processing end, and the user information push end is in communication connection with the user value model processing end and the user indicator data processing end respectively; and after the factor selection processing is performed on the initial user association factor information sample set, the initial value flow factor information sample set, the initial value type factor information sample set and the initial user portrait factor information sample set respectively, the method further includes: The user indicator data processing end sends the user association factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set to the user value model processing end.

4. The method according to claim 3, wherein: The method of training the initial user association information state recognition model, the initial value circulation information state recognition model, the initial value type information state recognition model and the initial user portrait state recognition model according to the user association factor information sample set, the value circulation factor information sample set, the value type factor information sample set and the user portrait factor information sample set respectively, and obtaining the trained user association information state recognition model, the value circulation information state recognition model, the value type information state recognition model and the user portrait state recognition model, including: Selecting a target user correlation factor information sample from the user correlation factor information sample set; Inputting the target user association factor information sample into the initial user association information state recognition model to obtain an initial user association information state recognition result, wherein the initial user association information state recognition result represents a state score of the user association factor information sample; Performing interpolation transformation on the initial user association information state recognition result to obtain initial user association information transformation interpolation; Determine an association difference value between the initial user association information transformation interpolation and the corresponding user association factor information sample label; In response to determining that the association difference value is less than or equal to the preset difference value, the initial user association information state recognition model is determined as the trained user association information state recognition model.

5. The method according to claim 1, wherein: The step of inputting the user information to be identified into the trained user value information identification model to obtain the user value information identification result includes: Inputting the user information to be identified into the user-related information state identification model included in the user value information identification model to obtain a user-related information state identification result; Inputting the user information to be identified into the value flow information state identification model included in the user value information identification model to obtain a value flow information state identification result; Inputting the user information to be identified into the value type information state identification model included in the user value information identification model to obtain a value type information state identification result; Inputting the user information to be identified into the user portrait state identification model included in the user value information identification model to obtain a user portrait state identification result; A user value information identification result is generated based on the user association information status identification result, the value flow information status identification result, the value type information status identification result and the user portrait status identification result.

6. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

7. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images