Search

CN-121997223-A - Method, device, equipment and storage medium for screening and identifying key users

CN121997223ACN 121997223 ACN121997223 ACN 121997223ACN-121997223-A

Abstract

The application belongs to the technical field of artificial intelligence, and relates to a key user screening and identifying method, device, equipment and storage medium, wherein multimode service data provided by a target user are obtained; the method comprises the steps of carrying out time sequence feature engineering processing to obtain dynamic index data introducing service time nodes, inputting the dynamic index data into a user screening and identifying agent, obtaining user category information, and identifying whether a target user is a key user according to the user category information. For example, when insurance contract signing and claim settlement business auditing are carried out, risk users in business users can be identified through the key user screening identification method, higher business risks are avoided, and moreover, the user screening identification intelligent body completed through integrated learning is adopted, and an artificial intelligence automatic processing mode is adopted, so that the key user identification efficiency is improved, the business identification misjudgment rate caused by only identifying key users through artificial experience is reduced to a certain extent, and the key user identification accuracy is improved.

Inventors

  • YUAN HUAN

Assignees

  • 中国平安财产保险股份有限公司

Dates

Publication Date
20260508
Application Date
20260113

Claims (10)

  1. 1. The key user screening and identifying method is characterized by comprising the following steps: acquiring multi-mode service data provided by a target user; Performing time sequence feature engineering processing on the multi-mode service data to obtain dynamic index data introducing service time nodes; Inputting the dynamic index data into a user screening and identifying agent with integrated learning; Acquiring user category information output by the user screening identification agent; And identifying whether the target user is a key user according to the user category information.
  2. 2. The focused user screening recognition method of claim 1, wherein prior to performing the step of inputting the dynamic index data into the ensemble-learning-completed user screening recognition agent, the method further comprises: Acquiring multi-mode business data of historical batch users, and constructing a training data set used for integrated learning; Performing pre-classification processing on the training data set according to the user classification labels respectively corresponding to each user to obtain the pre-classification class number and the pre-classification processing result of the training data set; inputting the training data set into an integrated learning model based on an improved AdaBoost algorithm, wherein the improved AdaBoost algorithm is computationally fused with a gradient lifting tree and a logistic regression algorithm; Setting user category output nodes at an output layer of the integrated learning model according to the pre-category number of the training data set; adopting an improved AdaBoost algorithm in the integrated learning model to carry out multiple rounds of iterative classification on training samples in the training data set to obtain an actual classification result of each round; sequentially comparing the difference between the actual classification dividing result of each round and the pre-classification processing result; And when the difference meets the preset integrated learning completion condition, obtaining the user screening and identifying intelligent agent with integrated learning completion.
  3. 3. The method for screening and identifying key users according to claim 2, wherein the step of setting user category output nodes at the output layer of the integrated learning model according to the number of pre-category categories of the training data set specifically comprises: Counting the number of pre-classified categories of the training data set; and setting equal user category output nodes at the output layer of the integrated learning model according to the pre-category number.
  4. 4. The method according to claim 2, wherein the step of sequentially comparing the difference between the actual classification division result and the pre-classification processing result of each round comprises: Acquiring an actual classification and division result of a current wheel; Sampling mode is adopted to extract a plurality of groups of pre-classification results from the pre-classification processing results; identifying whether the plurality of groups of pre-classification results all belong to a classification subset of the actual classification results of the current round; If the pre-classification results in the groups of pre-classification results do not belong to the classification subset of the actual classification division results of the current round, the preset integrated learning completion condition is not met, and the actual classification division result acquisition and the difference comparison of the next round are continued; and when the difference meets a preset integrated learning completion condition, obtaining the user screening and identifying the intelligent agent after the integrated learning is completed, wherein the method comprises the following steps of: If the plurality of groups of pre-classification results belong to the classification subsets of the actual classification dividing results of the current wheel, the preset integrated learning completion condition is met; And obtaining classification processing parameters corresponding to actual classification dividing results of the current round, and setting a strong classifier corresponding to the classification processing parameters as the user screening and identifying intelligent agent after the integrated learning is completed.
  5. 5. The method according to claim 2, wherein the step of sequentially comparing the difference between the actual classification division result and the pre-classification processing result of each round comprises: Acquiring an actual classification and division result of a current wheel; Identifying the consistency of the actual classification dividing result of the current wheel and the pre-classification processing result by adopting a full-quantity comparison mode; If the consistency does not exceed the preset similarity threshold, the preset integrated learning completion condition is not met, and the acquisition of the actual classification result and the difference comparison of the next round are continued; and when the difference meets a preset integrated learning completion condition, obtaining the user screening and identifying the intelligent agent after the integrated learning is completed, wherein the method comprises the following steps of: if the consistency exceeds a preset similarity threshold, a preset integrated learning completion condition is met; And obtaining classification processing parameters corresponding to actual classification dividing results of the current round, and setting a strong classifier corresponding to the classification processing parameters as the user screening and identifying intelligent agent after the integrated learning is completed.
  6. 6. The focused user screening and identifying method according to claim 2, wherein after the step of obtaining the user screening and identifying agent for completion of the ensemble learning when the variability satisfies a preset ensemble learning completion condition is performed, the method further comprises: respectively acquiring the dynamic index data of the user service data in all classification categories in the pre-classification processing result; Determining differential dynamic index data corresponding to all classification categories respectively through comprehensive comparison; And constructing a mapping relation between the differentiated dynamic index data and the user category output nodes according to the differentiated dynamic index data respectively corresponding to all the classification categories and the user category output nodes respectively corresponding to all the classification categories.
  7. 7. The focused user screening recognition method of claim 6, wherein after performing the step of inputting the dynamic index data into the ensemble-learning-completed user screening recognition agent, the method further comprises: Identifying differentiated dynamic index data contained in the dynamic index data; Screening out corresponding user class output nodes as user class target output nodes according to the mapping relation between the differentiated dynamic index data and the user class output nodes; And determining the user category information according to the output result of the user category target output node.
  8. 8. An accent user screening and identifying device, comprising: The service data acquisition module is used for acquiring multi-mode service data provided by a target user; The dynamic index data acquisition module is used for carrying out time sequence characteristic engineering processing on the multi-mode service data to obtain dynamic index data introduced with service time nodes; The dynamic index data input module is used for inputting the dynamic index data into the user screening and identifying agent with integrated learning; the user category information output and acquisition module is used for acquiring user category information output by the user screening and identifying intelligent agent; And the key user identification module is used for identifying whether the target user is a key user or not according to the user category information.
  9. 9. A computer device comprising a memory having stored therein computer readable instructions which when executed by the processor implement the steps of the focused user selection identification method of any one of claims 1 to 7.
  10. 10. A computer readable storage medium having stored thereon computer readable instructions which when executed by a processor implement the steps of the focused user selection identification method of any of claims 1 to 7.

Description

Method, device, equipment and storage medium for screening and identifying key users Technical Field The application relates to the technical field of artificial intelligence, which is applied to a scene of screening and identifying important users when target business signing is carried out, and relates to a screening and identifying method, device, equipment and storage medium for the important users. Background With the rapid development of artificial intelligence technology, the intelligent technology is widely applied in various fields of various lines, such as automatic processing of financial services, and more processing of financial services can be realized through the internet. To ensure the security and emphasis of financial business processing using the internet. It is desirable to identify key users in a user group, such as financial service users with good credit conditions, financial service users with full asset condition. In the prior art, the method for checking the heavy-point user is that after the user submits the service request, the service request data is checked and identified manually. The key users are easily influenced by subjective factors to cause low recognition accuracy through manual recognition, and the time consumption is long, so that the recognition speed is low. Disclosure of Invention The embodiment of the application aims to provide a key user screening and identifying method, device, equipment and storage medium, so as to improve the key user identifying efficiency and accuracy. In a first aspect, an embodiment of the present application provides a method for screening and identifying key users, which adopts the following technical schemes: a key user screening and identifying method comprises the following steps: acquiring multi-mode service data provided by a target user; Performing time sequence feature engineering processing on the multi-mode service data to obtain dynamic index data introducing service time nodes; Inputting the dynamic index data into a user screening and identifying agent with integrated learning; Acquiring user category information output by the user screening identification agent; And identifying whether the target user is a key user according to the user category information. In a second aspect, an embodiment of the present application further provides a key user screening and identifying device, which adopts the following technical scheme: an accent user screening and identifying device, comprising: The service data acquisition module is used for acquiring multi-mode service data provided by a target user; The dynamic index data acquisition module is used for carrying out time sequence characteristic engineering processing on the multi-mode service data to obtain dynamic index data introduced with service time nodes; The dynamic index data input module is used for inputting the dynamic index data into the user screening and identifying agent with integrated learning; the user category information output and acquisition module is used for acquiring user category information output by the user screening and identifying intelligent agent; And the key user identification module is used for identifying whether the target user is a key user or not according to the user category information. In a third aspect, an embodiment of the present application further provides a computer device, which adopts the following technical scheme: A computer device comprising a memory and a processor, wherein computer readable instructions are stored in the memory, and the processor when executing the computer readable instructions implements the steps of the above-described focused user screening identification method. In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, which adopts the following technical solutions: a computer readable storage medium having stored thereon computer readable instructions which when executed by a processor perform the steps of the focused user selection identification method as described above. Compared with the prior art, the embodiment of the application has the following main beneficial effects: The key user screening and identifying method can be widely applied to a scene of screening and identifying key users when target service signing is carried out, multi-mode service data provided by the target users are obtained, time sequence characteristic engineering processing is carried out, dynamic index data introducing service time nodes is obtained, the dynamic index data are input into a user screening and identifying intelligent body, user category information is obtained, and whether the target users are key users or not is identified according to the user category information. For example, when insurance contract signing and claim settlement business auditing are carried out, the key user screening and identifying method can assist financial businesses such as insur