Artificial intelligence-based crowd identification method and device, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to a crowd recognition method and device based on artificial intelligence, computer equipment and a storage medium. Wherein the portrait data comprises basic attribute data, value attribute data and behavior attribute data; performing feature engineering processing on the portrait data to obtain first portrait data; performing standardization processing on the first portrait data to obtain second portrait data; performing clustering processing on the second portrait data to obtain a clustering result; screening out high-value crowds from all agents based on a clustering result; generating a loss probability of the high-value crowd based on a preset target classification model; and screening out high-value and high-risk loss crowds from the high-value crowds based on the loss probability. In addition, the churn probability may be stored in a blockchain. According to the method, based on the use of the clustering algorithm and the target classification model, the high-value and high-risk loss crowd in the agents can be automatically and accurately identified.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to methods, devices, computer equipment and storage media for crowd identification based on artificial intelligence. Background Art

[0002] In the insurance industry, insurance agents are an important bridge between insurance companies and customers. Their stability and performance are directly related to the company's business development and market competitiveness. However, in recent years, with the changes in the market environment, the intensification of industry competition and the diversification of agents' personal career plans, the problem of insurance agent turnover has become increasingly prominent, especially the high turnover rate of new agents within one year, and the increasing turnover rate of long-term agents with high service years, which has become one of the key factors restricting the healthy development of the industry.

[0003] In response to this situation, insurance companies and industry managers have been exploring effective response strategies in order to improve the retention rate and performance contribution of agents through refined management. Among them, the segmentation of agent groups based on business logic has become an important management tool. Through in-depth analysis of multi-dimensional data such as agents' business performance, work habits, personal characteristics, etc., agents are divided into different levels or groups, such as top performers (MVP), high performers (QVP) and quasi-performers (FVP), etc., in order to more accurately identify the development needs and potential problems of agents, so as to implement differentiated management strategies and incentives. However, although the above classification method has improved the pertinence and effectiveness of management to a certain extent, in actual operation, due to the lack of an effective early identification mechanism, intervention can only be carried out after the problem appears, which greatly limits the timeliness and effectiveness of the assistance measures.

[0004] At present, in the field of insurance agent loss warning, existing technologies still make it difficult to achieve early identification of high-risk loss groups. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a method, device, computer equipment and storage medium for identifying people based on artificial intelligence, so as to solve the technical problem that in the field of insurance agent loss warning, the existing technology is still difficult to achieve early identification of high-risk loss people.

[0006] In order to solve the above technical problems, the present application embodiment provides a crowd recognition method based on artificial intelligence, which adopts the following technical solutions:

[0007] Acquire agent portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data;

[0008] Performing feature engineering processing on the portrait data to obtain corresponding first portrait data;

[0009] Performing standardization processing on the first portrait data to obtain corresponding second portrait data;

[0010] Performing clustering processing on the second portrait data to obtain corresponding clustering results;

[0011] Screening out high-value people from all the agents based on the clustering results;

[0012] Generate the churn probability of the high-value population based on a preset target classification model;

[0013] Based on the loss probability, a high-value and high-risk loss population is screened out from the high-value population.

[0014] Furthermore, the step of performing feature engineering processing on the portrait data to obtain corresponding first portrait data specifically includes:

[0015] Performing missing value filling processing on the portrait data to obtain corresponding first processed data;

[0016] Encoding the first processed data to obtain corresponding second processed data;

[0017] performing outlier processing on the second processed data to obtain corresponding third processed data;

[0018] Performing multicollinearity elimination processing on the third processed data to obtain corresponding fourth processed data;

[0019] The fourth processed data is used as the first portrait data.

[0020] Furthermore, the step of performing clustering processing on the second portrait data to obtain a corresponding clustering result specifically includes:

[0021] Call the target clustering algorithm;

[0022] Determining the number of cluster centers corresponding to the target clustering algorithm;

[0023] Based on the number of cluster centers, the target clustering algorithm is used to perform clustering processing on the second portrait data to obtain the corresponding clustering result.

[0024] Furthermore, the step of screening out high-value people from all the agents based on the clustering result specifically includes:

[0025] For each of the clustering results, obtaining the specified value attribute data of each of the clustering results;

[0026] Generating a value attribute mean value of each clustering result based on the designated value attribute data of each clustering result;

[0027] Sort all the clustering results by the mean values ​​of the value attributes in descending order to obtain corresponding clustering sorting results;

[0028] Filter out the designated clustering result ranked first from the clustering sorting results;

[0029] The designated clustering result is used as the high-value population.

[0030] Furthermore, the step of generating the churn probability of the high-value population based on the preset target classification model specifically includes:

[0031] Acquire agent data corresponding to the high-value population;

[0032] Inputting the agent data into the target classification model, and outputting the decision value corresponding to the high-value population through the target classification model;

[0033] Call the preset conversion function;

[0034] Convert the decision value based on the conversion function to obtain a corresponding probability value;

[0035] The probability value is used as the loss probability of the high-value population.

[0036] Furthermore, the step of screening out a high-value and high-risk churn population from the high-value population based on the churn probability specifically includes:

[0037] Obtain the high-risk churn threshold range corresponding to the preset high-value high-risk churn population;

[0038] Obtaining the first churn probability of all first agents included in the high-value population;

[0039] Filter out a second churn probability that meets the high-risk churn threshold range from all the first churn probabilities;

[0040] Selecting a second agent corresponding to the second churn probability from all the first agents;

[0041] The high-value and high-risk loss population is generated based on the second agent.

[0042] Furthermore, after the step of screening out high-value and high-risk churn population from the high-value population based on the churn probability, the method further includes:

[0043] Based on the preset information type, target information corresponding to high-value and high-risk loss groups is obtained;

[0044] Generate corresponding warning information based on the target information;

[0045] Obtain the preset retention strategy content;

[0046] Determine the target notification method;

[0047] Based on the target notification method, the target information, the warning information and the retention strategy content are sent to relevant personnel.

[0048] In order to solve the above technical problems, the embodiment of the present application further provides a crowd recognition device based on artificial intelligence, which adopts the following technical solution:

[0049] A first acquisition module is used to acquire the agent's portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data;

[0050] A first processing module, used for performing feature engineering processing on the portrait data to obtain corresponding first portrait data;

[0051] A second processing module, used for performing standardization processing on the first portrait data to obtain corresponding second portrait data;

[0052] A clustering module, used for performing clustering processing on the second portrait data to obtain corresponding clustering results;

[0053] A first screening module, used for screening out high-value people from all the agents based on the clustering result;

[0054] A first generating module, used to generate the churn probability of the high-value population based on a preset target classification model;

[0055] The second screening module is used to screen out high-value and high-risk churn population from the high-value population based on the churn probability.

[0056] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0057] Acquire agent portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data;

[0058] Performing feature engineering processing on the portrait data to obtain corresponding first portrait data;

[0059] Performing standardization processing on the first portrait data to obtain corresponding second portrait data;

[0060] Performing clustering processing on the second portrait data to obtain corresponding clustering results;

[0061] Screening out high-value people from all the agents based on the clustering results;

[0062] Generate the churn probability of the high-value population based on a preset target classification model;

[0063] Based on the loss probability, a high-value and high-risk loss population is screened out from the high-value population.

[0064] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0065] Acquire agent portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data;

[0066] Performing feature engineering processing on the portrait data to obtain corresponding first portrait data;

[0067] Performing standardization processing on the first portrait data to obtain corresponding second portrait data;

[0068] Performing clustering processing on the second portrait data to obtain corresponding clustering results;

[0069] Screening out high-value people from all the agents based on the clustering results;

[0070] Generate the churn probability of the high-value population based on a preset target classification model;

[0071] Based on the loss probability, a high-value and high-risk loss population is screened out from the high-value population.

[0072] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0073] The present application first obtains the agent's portrait data; wherein the portrait data includes at least basic attribute data, value attribute data and behavioral attribute data; then the portrait data is subjected to feature engineering processing to obtain corresponding first portrait data; then the first portrait data is subjected to standardization processing to obtain corresponding second portrait data; subsequently the second portrait data is subjected to clustering processing to obtain corresponding clustering results, and high-value people are screened out from all the agents based on the clustering results; further the loss probability of the high-value people is generated based on a preset target classification model; finally, high-value and high-risk loss people are screened out from the high-value people based on the loss probability. The present application obtains second portrait data by performing feature engineering and standardization processing on the acquired agent portrait data, then performs clustering processing on the second portrait data and screens out high-value people from all the agents based on the obtained clustering results, and subsequently generates the loss probability of the high-value people based on a preset target classification model, and then screens out high-value and high-risk loss people from the high-value people based on the loss probability, so as to automatically and accurately identify the loss risk of high-value and high-risk loss people among agents, improve the intelligent identification of high-risk loss people, and facilitate the subsequent timely adoption of corresponding measures for early warning response, thereby reducing the loss rate of high-value and high-risk loss people. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0075] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0076] Figure 2 A flow chart of an embodiment of a crowd recognition method based on artificial intelligence according to the present application;

[0077] Figure 3 is a structural schematic diagram of an embodiment of an artificial intelligence-based crowd recognition device according to the present application;

[0078] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0079] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0080] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0081] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0082] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0083] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0084] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0085] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0086] It should be noted that the artificial intelligence-based crowd identification method provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based crowd identification device is generally set in the server / terminal device.

[0087] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0088] Continue to refer Figure 2 , showing a flow chart of an embodiment of an artificial intelligence-based crowd identification method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The artificial intelligence-based crowd identification method provided in the embodiment of the present application can be applied to any scenario where the loss of people is required, and the artificial intelligence-based crowd identification method can be applied to products in these scenarios, for example, the loss of people in the financial insurance field. The artificial intelligence-based crowd identification method includes the following steps:

[0089] Step S201, obtaining the agent's portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data.

[0090] In this embodiment, the electronic device (e.g. Figure 1The server / terminal device shown in the figure) can obtain the agent's portrait data through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The present application can be applied to business scenarios in the field of financial insurance to identify the risk of loss of high-value people among insurance agents. The above-mentioned agents are specifically insurance agents in insurance companies. In order to more comprehensively portray insurance agents and combine the dimensions of frequent business analysis, the present application establishes an agent's portrait by using three dimensions: basic attributes, value attributes, and behavioral attributes, wherein the basic attribute data includes the agent's own basic labels (such as gender, age, company length, etc.), the value attribute data includes the agent's performance indicators (such as the number of long-term insurance cases, premium income, and other performance indicators), and the behavioral attributes include the agent's activity indicators (such as attendance rate, number of customer visits, and other activity indicators).

[0091] Step S202, performing feature engineering processing on the portrait data to obtain corresponding first portrait data.

[0092] In the present embodiment, the above-mentioned specific implementation process of performing feature engineering on the portrait data to obtain the corresponding first portrait data will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated in detail here.

[0093] Step S203: standardize the first portrait data to obtain corresponding second portrait data.

[0094] In this embodiment, after the portrait data is subjected to feature engineering to obtain the first portrait data, in order to eliminate the influence of the dimensions and magnitudes between the features contained in the first portrait data so that the features can be compared and weighted, the first portrait data can be further standardized. Among them, the standardization can be performed by using the formula X^=(X-μ) / σ. X is the original data, μ is the data mean, σ is the data standard deviation, and X^ is the standardized data.

[0095] Step S204, clustering the second portrait data to obtain corresponding clustering results.

[0096] In this embodiment, the above-mentioned specific implementation process of clustering the second portrait data to obtain the corresponding clustering results will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated in detail here.

[0097] Step S205 , screening out high-value people from all the agents based on the clustering result.

[0098] In this embodiment, the specific implementation process of screening out high-value people from all the agents based on the clustering results will be described in further detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0099] Step S206: generating the churn probability of the high-value population based on a preset target classification model.

[0100] In this embodiment, the specific implementation process of generating the loss probability of the high-value population based on the preset target classification model will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0101] Step S207, screening out high-value and high-risk churn population from the high-value population based on the churn probability.

[0102] In this embodiment, the specific implementation process of screening out the high-value and high-risk loss population from the high-value population based on the loss probability will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0103] The present application first obtains the agent's portrait data; wherein the portrait data includes at least basic attribute data, value attribute data and behavioral attribute data; then the portrait data is subjected to feature engineering processing to obtain corresponding first portrait data; then the first portrait data is subjected to standardization processing to obtain corresponding second portrait data; subsequently the second portrait data is subjected to clustering processing to obtain corresponding clustering results, and high-value people are screened out from all the agents based on the clustering results; further the loss probability of the high-value people is generated based on a preset target classification model; finally, high-value and high-risk loss people are screened out from the high-value people based on the loss probability. The present application obtains second portrait data by performing feature engineering and standardization processing on the acquired agent portrait data, then performs clustering processing on the second portrait data and screens out high-value people from all the agents based on the obtained clustering results, and subsequently generates the loss probability of the high-value people based on a preset target classification model, and then screens out high-value and high-risk loss people from the high-value people based on the loss probability, so as to automatically and accurately identify the loss risk of high-value and high-risk loss people among agents, improve the intelligent identification of high-risk loss people, and facilitate the subsequent timely adoption of corresponding measures for early warning response, thereby reducing the loss rate of high-value and high-risk loss people.

[0104] In some optional implementations, step S202 includes the following steps:

[0105] Perform missing value filling processing on the portrait data to obtain corresponding first processed data.

[0106] In this embodiment, the missing value filling process may include filling the missing values ​​in the portrait data with default values, for example, the number of long-term insurance cases is set to 0 by default.

[0107] The first processed data is encoded to obtain corresponding second processed data.

[0108] In this embodiment, an unsupervised encoding method may be used to encode the variable data in the first processed data, such as age, gender, etc., so as to obtain the corresponding second processed data. There is no specific limitation on the selection of the above unsupervised encoding method, for example, any one of one-hot encoding, dummy variable encoding, and label encoding may be used.

[0109] Perform outlier processing on the second processed data to obtain corresponding third processed data.

[0110] In this embodiment, the processing of outliers in the second processed data follows the 3σ principle (Laida criterion) in statistics, that is, normal values ​​are almost all concentrated in the interval (μ-3σ, μ+3σ), and the possibility of exceeding this range is less than 0.3%. They are considered to be outliers and can be filled with the 3σ boundary value.

[0111] The third processed data is processed to eliminate multicollinearity to obtain corresponding fourth processed data.

[0112] In this embodiment, in order to address the possible multicollinearity between the various features contained in the third processed data (i.e., the phenomenon that the characteristics are correlated with each other, which will lead to unstable analysis results when serious collinearity problems occur), a correlation analysis is performed between the features. When the correlation coefficient is higher than 0.8, it indicates that multicollinearity exists between the features, and then it is necessary to eliminate multicollinearity by removing collinear features, principal component analysis, and other methods.

[0113] The fourth processed data is used as the first portrait data.

[0114] The present application performs missing value filling processing on the portrait data to obtain the corresponding first processed data; then performs encoding processing on the first processed data to obtain the corresponding second processed data; then performs outlier processing on the second processed data to obtain the corresponding third processed data; subsequently performs multicollinearity elimination processing on the third processed data to obtain the corresponding fourth processed data; finally, the fourth processed data is used as the first portrait data. The present application performs missing value filling processing, encoding processing, outlier processing, and multicollinearity elimination processing on the acquired agent's portrait data, thereby realizing intelligent and accurate feature engineering processing for the portrait data, effectively improving the processing efficiency of the feature engineering processing, and ensuring the data accuracy of the obtained first portrait data.

[0115] In some optional implementations of this embodiment, step S204 includes the following steps:

[0116] Invoke the target clustering algorithm.

[0117] In this embodiment, the target clustering algorithm may specifically adopt the K-means clustering algorithm. The principle of the K-means clustering algorithm is as follows: 1. Randomly select k points as the initial cluster centers; 2. For the remaining points, classify them into the nearest cluster according to their distance from the cluster center; 3. For each cluster, calculate the mean of all points as the new cluster center; repeat steps 2 and 3 until the cluster center no longer changes.

[0118] The number of cluster centers corresponding to the target clustering algorithm is determined.

[0119] In this embodiment, in order to divide the population into low-value population, medium-value population and high-value population, the number of cluster centers corresponding to the target clustering algorithm is set to 3, that is, K is 3.

[0120] Based on the number of cluster centers, the target clustering algorithm is used to perform clustering processing on the second portrait data to obtain the corresponding clustering result.

[0121] In this embodiment, the target features can be obtained by selecting features directly related to the agent value assessment from the second portrait data, such as premium income, number of long-term insurance cases, etc. Then, K-means clustering processing is performed based on the target features. Specifically, K-means clustering processing includes initialization and iteration processes. Initialization: Select K=3 as the number of cluster centers, and select the initial cluster center randomly or according to a certain strategy (such as K-means++). Iteration process: 1. Allocation step: For each agent in the data set, assign it to the nearest cluster center according to its distance from each cluster center. 2. Update step: Recalculate the center of each cluster (that is, calculate the mean of the value attributes of all agents in the cluster). 3. Repeat the allocation step and the update step until the cluster center no longer changes significantly or reaches the preset number of iterations, thereby obtaining the corresponding clustering result. In addition, the clustering results can also be evaluated to ensure the accuracy of the clustering results. Specifically, it can be checked whether the clustering converges, that is, whether the cluster center is stable during the iteration process. Alternatively, the clustering effect can be evaluated by indicators such as the Silhouette Coefficient.

[0122] The present application calls a target clustering algorithm, then determines the number of cluster centers corresponding to the target clustering algorithm, and then uses the target clustering algorithm to cluster the second portrait data based on the number of cluster centers to obtain the corresponding clustering result. The present application can quickly and accurately generate corresponding clustering results by using a target clustering algorithm and clustering the second portrait data based on the determined number of cluster centers corresponding to the target clustering algorithm, thereby ensuring the accuracy of the obtained clustering results, which is conducive to accurately screening out high-value people from all the agents based on the obtained clustering results, thereby improving the screening accuracy of high-value people.

[0123] In some optional implementations, step S205 includes the following steps:

[0124] For each of the clustering results, the designated value attribute data of each of the clustering results is obtained.

[0125] In this embodiment, the designated value attribute data of the clustering result refers to all the value attribute data corresponding to all the agents included in the clustering result.

[0126] The value attribute mean of each of the clustering results is generated based on the designated value attribute data of each of the clustering results.

[0127] In this embodiment, for each clustering result obtained through clustering processing, the average value of each value attribute of all agents in the clustering result is calculated, and then these average values ​​are weighted averaged to obtain the value attribute mean of each clustering result. The weight of each attribute can be set based on the different importance of the attributes.

[0128] All the clustering results are sorted in descending order according to the mean values ​​of the value attributes to obtain corresponding clustering sorting results.

[0129] In this embodiment, by numerically comparing the value attribute means of each clustering result, all the clustering results can be sorted in descending order according to the value attribute means, thereby obtaining a corresponding clustering sorting result.

[0130] A designated clustering result ranked first is selected from the clustering sorting results.

[0131] In this embodiment, after all clustering results are sorted according to the calculated value attribute mean values, the clustering result with the highest value attribute mean value is taken as the high-value population.

[0132] The designated clustering result is used as the high-value population.

[0133] For each of the clustering results, the present application obtains the designated value attribute data of each of the clustering results; then, based on the designated value attribute data of each of the clustering results, the value attribute mean of each of the clustering results is generated; then, the value attribute mean is used to sort all of the clustering results in order from high to low to obtain the corresponding clustering sorting result; subsequently, the designated clustering result ranked first is selected from the clustering sorting result, and the designated clustering result is used as the high-value population. The present application calculates the mean of the value attribute for each of the clustering results to obtain the value attribute mean of each clustering result, then, the value attribute mean is used to sort all of the clustering results in order from high to low to obtain the corresponding clustering sorting result, and then, the designated clustering result ranked first is selected from the clustering sorting result as the final high-value population, so as to realize fast and intelligent screening of high-value population from all of the agents based on the clustering results, thereby improving the screening intelligence of high-value population and ensuring the accuracy of the obtained high-value population.

[0134] In some optional implementations, step S206 includes the following steps:

[0135] The agent data corresponding to the high-value population is obtained.

[0136] In this embodiment, agent data classified as high-value people is obtained, and the agent data at least includes data corresponding to the value attributes and behavior attributes of the agents that are helpful for predicting the churn probability.

[0137] The agent data is input into the target classification model, and the decision value corresponding to the high-value population is output through the target classification model.

[0138] In this embodiment, the target classification model is specifically a model obtained by training the SVM classification model using pre-collected sample data. The decision value is usually the distance from the hyperplane or a symbolic category decision, or is called a category label (churn or employment).

[0139] Specifically, the construction process of the target classification model includes: first, obtaining the high-value population data corresponding to the high-value population in the historical time period. The high-value population data at least includes data such as the value attributes and behavioral attributes of the agent. Then, for the high-value population data, a label is assigned to each agent (y=1 for churn, y=0 for in-service). These labels usually come from historical data, such as the company's records of agent churn. Then, the SVM model is trained using the labeled high-value population data. Including: 1) Perform necessary preprocessing on the extracted high-value population data, such as standardization or normalization, to ensure that each feature has the same weight in the model. Select the SVM kernel function. 2) Select a suitable SVM kernel function (such as linear kernel, polynomial kernel, RBF kernel, etc.), which depends on the characteristics of the data and the complexity of the problem. For high-dimensional data or nonlinear relationships, the RBF (radial basis function) kernel is usually a good choice. 3) Use the labeled high-value population data to train the SVM model. During the training process, the SVM will try to find a hyperplane in the feature space that can maximize the interval between the two categories (churn and in-service). Then, SVM finds this optimal hyperplane during the training process through an optimization algorithm (such as SMO, sequential minimum optimization algorithm). The position of the hyperplane is determined by the support vector (that is, the data points on the boundaries of both sides of the hyperplane). The SVM model is subsequently evaluated, including: 1) Cross-validation. Use cross-validation (such as k-fold cross-validation) to evaluate the accuracy and robustness of the model. Divide the data set into k subsets, use k-1 subsets for training each time, and the remaining subset for testing. Repeat k times, each time selecting a different subset as the test set, and finally calculate the average performance indicators (such as accuracy, recall, F1 score, etc.). 2) Performance evaluation. Based on the results of cross-validation, evaluate whether the performance of the model meets the business requirements. If the performance is poor, you may need to adjust the model parameters (such as C parameters, kernel function parameters, etc.), or reconsider the feature selection and data preprocessing steps.

[0140] Call the preset conversion function.

[0141] In this embodiment, the above conversion function may specifically adopt a Sigmoid function. The Sigmoid function may be used to map the output (ie, decision value) of the target classification model to a (0, 1) interval, which may be interpreted as the probability of churn.

[0142] The decision value is converted based on the conversion function to obtain a corresponding probability value.

[0143] In this embodiment, the above conversion function may specifically be: Where P(y=1|x i ) is a conditional probability, indicating that given the agent feature x i The probability of the agent leaving (y = 1) under the condition of i ) is the target classification model for agent x i The decision function outputs the value. In the standard SVM model, this value is usually the agent's feature vector x i The signed distance to the decision hyperplane (also called the decision value or score). This value can be positive, negative, or zero, corresponding to different class decisions (for example, a positive number represents class 1 and a negative number represents class 0). However, in order to convert it to a probability, we need to perform a further transformation on this value. The two parameters A and B are the parameters of the Sigmoid function (also called the logistic function) used to convert the decision value f(x) of the SVM model to i ) is mapped to the interval (0,1) to obtain the churn probability. These parameters are usually not obtained directly through SVM model training, but need to be estimated through additional calibration processes (such as Platt's Scaling method). A (also called slope or scaling factor) controls the slope of the Sigmoid function, that is, it determines f(x i ) values ​​are converted nonlinearly to probability values. B (also called offset or intercept) is used to adjust the horizontal position of the Sigmoid function to ensure that when f(x i ) is a certain value, the obtained probability value is close to 0.5 (i.e., the decision boundary between the two categories). The calibration process usually involves the following steps: Use an independent calibration set (usually a subset of the training set or a completely independent validation set) that contains samples of known categories. For each sample in the calibration set, use the SVM model to calculate its decision value f(x i ). Use logistic regression (or other calibration methods) to fit the relationship between these decision values ​​and the corresponding class labels to estimate the values ​​of A and B.

[0144] The probability value is used as the loss probability of the high-value population.

[0145] The present application obtains agent data corresponding to the high-value population; then inputs the agent data into the target classification model, and outputs the decision value corresponding to the high-value population through the target classification model; then calls a preset conversion function; subsequently converts the decision value based on the conversion function to obtain a corresponding probability value, and uses the probability value as the loss probability of the high-value population. The present application obtains agent data corresponding to the high-value population, and then processes the agent data of the high-value population based on the use of the target classification model and the conversion function, so that the loss probability of the high-value population can be automatically and accurately generated, thereby improving the generation efficiency of the loss probability of the high-value population and ensuring the numerical accuracy of the obtained loss probability.

[0146] In some optional implementations of this embodiment, step S207 includes the following steps:

[0147] Get the high-risk churn threshold range corresponding to the preset high-value and high-risk churn population.

[0148] In this embodiment, a lower low-risk churn threshold range is predetermined as the boundary of low-risk churn. This threshold can be set based on historical data, business needs, and the acceptable churn rate of the company. For example, p∈[0%,50%) can be set as a high-value, low-risk churn population. Then a medium medium-risk churn threshold range is determined to distinguish between medium-risk churn and high-risk churn. This threshold also needs to be set based on historical data and business needs. For example, p∈[50%,75%) can be set as a high-value medium-risk churn population. Finally, a higher high-risk churn threshold range is determined to identify high-risk churn populations. This threshold should be able to accurately capture those high-value populations that are most likely to churn. For example, p∈[75%,100%] can be set as a high-value, high-risk churn population.

[0149] The first churn probabilities of all first agents included in the high-value population are obtained.

[0150] In this embodiment, information extraction may be performed on the high-value population to obtain the first churn probabilities of all first agents included in the high-value population.

[0151] A second churn probability that meets the high-risk churn threshold range is screened out from all the first churn probabilities.

[0152] In this embodiment, the second loss probability refers to data whose probability value is within the high-risk loss threshold range.

[0153] A second agent corresponding to the second churn probability is selected from all the first agents.

[0154] In this embodiment, the second agent refers to an agent among all the first agents that has a matching relationship with the second churn probability.

[0155] The high-value and high-risk loss population is generated based on the second agent.

[0156] In this embodiment, all the screened second agents may be integrated to obtain integrated agents, and the integrated agents may be used as the high-value and high-risk loss population.

[0157] The present application obtains a high-risk loss threshold range corresponding to a preset high-value, high-risk loss population; then obtains the first loss probability of all first agents included in the high-value population; then screens out a second loss probability that meets the high-risk loss threshold range from all the first loss probabilities; subsequently screens out a second agent corresponding to the second loss probability from all the first agents; and finally generates the high-value, high-risk loss population based on the second agent. The present application obtains a high-risk loss threshold range corresponding to a preset high-value, high-risk loss population, and obtains the first loss probability of all first agents included in the high-value population, and then numerically matches the first loss probability with the high-risk loss threshold range to obtain a second loss probability that meets the high-risk loss threshold range. Subsequently, based on the second agent corresponding to the second loss probability, the corresponding high-value, high-risk loss population can be quickly and accurately generated, thereby automatically completing the process of screening out high-value, high-risk loss populations from the high-value population based on the loss probability, thereby improving the screening efficiency of high-value, high-risk loss populations and ensuring the data accuracy of the obtained high-value, high-risk loss populations.

[0158] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps:

[0159] Based on the preset information type, obtain target information corresponding to high-value and high-risk loss groups.

[0160] In this embodiment, the above information types may at least include basic information of the agent, loss probability, etc., or may also include possible risk factor types.

[0161] Generate corresponding warning information based on the target information.

[0162] In this embodiment, the target information can be filled into a preset warning information template to generate corresponding warning information. The warning information template can be written and generated according to actual warning requirements.

[0163] Get the preset retention strategy content.

[0164] In this embodiment, retention strategy content corresponding to the high-value and high-risk churn population is pre-formulated, and the retention strategy content includes targeted retention measures or intervention plans. Exemplarily, the retention strategy content may include: strengthening communication with the high-value and high-risk churn population, understanding their needs and dissatisfaction, and solving problems in a timely manner. Alternatively, providing personalized incentive programs, such as increasing commissions, providing training opportunities, etc., to enhance their loyalty and satisfaction. Alternatively, analyzing the reasons for churn and improving products or services to reduce future churn risks.

[0165] Determine the target notification method.

[0166] In this embodiment, there is no specific limitation on the selection method of the above-mentioned target notification method, which can be determined according to actual notification needs, for example, email, text message, internal system message, etc. can be used.

[0167] Based on the target notification method, the target information, the warning information and the retention strategy content are sent to relevant personnel.

[0168] In this embodiment, the above-mentioned relevant personnel may refer to management personnel of the human resources department.

[0169] This application obtains target information corresponding to the high-value and high-risk churn population based on a preset information type; then generates corresponding warning information based on the target information; then obtains preset retention strategy content; subsequently determines the target notification method; and finally, based on the target notification method, sends the target information, the warning information, and the retention strategy content to relevant personnel. After screening out the high-value and high-risk churn population from the high-value population based on the churn probability, this application will intelligently obtain target information corresponding to the high-value and high-risk churn population, and generate corresponding warning information based on the target information, and obtain preset retention strategy content, and then use the target notification method to send the target information, the warning information, and the retention strategy content to relevant personnel, so as to realize automatic and intelligent early warning processing for the high-value and high-risk churn population, timely discover and respond to the churn risk of the high-value and high-risk churn population, thereby reducing the churn rate of the high-value and high-risk churn population, and is conducive to improving customer satisfaction and loyalty.

[0170] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0171] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned loss probability, the above-mentioned loss probability can also be stored in a node of a blockchain.

[0172] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0173] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0174] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0175] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0176] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0177] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a crowd recognition device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0178] like Figure 3 As shown, the crowd identification device 300 based on artificial intelligence described in this embodiment includes: a first acquisition module 301, a first processing module 302, a second processing module 303, a clustering module 304, a first screening module 305, a first generation module 306 and a second screening module 307. Among them:

[0179] The first acquisition module 301 is used to acquire the agent's portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data;

[0180] A first processing module 302 is used to perform feature engineering processing on the portrait data to obtain corresponding first portrait data;

[0181] A second processing module 303 is used to perform standardization processing on the first portrait data to obtain corresponding second portrait data;

[0182] A clustering module 304 is used to perform clustering processing on the second portrait data to obtain corresponding clustering results;

[0183] A first screening module 305, for screening out high-value people from all the agents based on the clustering result;

[0184] A first generating module 306, configured to generate the churn probability of the high-value population based on a preset target classification model;

[0185] The second screening module 307 is used to screen out high-value and high-risk churn population from the high-value population based on the churn probability.

[0186] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0187] In some optional implementations of this embodiment, the first processing module 302 includes:

[0188] A first processing submodule is used to perform missing value filling processing on the portrait data to obtain corresponding first processed data;

[0189] A second processing submodule, used for encoding the first processed data to obtain corresponding second processed data;

[0190] A third processing submodule, configured to perform outlier processing on the second processed data to obtain corresponding third processed data;

[0191] A fourth processing submodule, used for performing multicollinearity elimination processing on the third processed data to obtain corresponding fourth processed data;

[0192] The first determining submodule is used to use the fourth processed data as the first portrait data.

[0193] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0194] In some optional implementations of this embodiment, the clustering module 304 includes:

[0195] The first calling submodule is used to call the target clustering algorithm;

[0196] A second determination submodule, used to determine the number of cluster centers corresponding to the target clustering algorithm;

[0197] A clustering submodule is used to perform clustering processing on the second portrait data based on the number of cluster centers using the target clustering algorithm to obtain the corresponding clustering result.

[0198] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0199] In some optional implementations of this embodiment, the first screening module 305 includes:

[0200] A first acquisition submodule, for acquiring, for each of the clustering results, designated value attribute data of each of the clustering results;

[0201] A first generating submodule, configured to generate a value attribute mean value of each clustering result based on the specified value attribute data of each clustering result;

[0202] A sorting submodule, used to sort all the clustering results according to the value attribute means from high to low, to obtain corresponding clustering sorting results;

[0203] A first screening submodule, used for screening out the first-ranked designated clustering result from the clustering sorting results;

[0204] The third determination submodule is used to use the designated clustering result as the high-value population.

[0205] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0206] In some optional implementations of this embodiment, the first generating module 306 includes:

[0207] The second acquisition submodule is used to acquire agent data corresponding to the high-value group;

[0208] An output submodule, used for inputting the agent data into the target classification model, and outputting the decision value corresponding to the high-value population through the target classification model;

[0209] The second calling submodule is used to call a preset conversion function;

[0210] A conversion submodule, used to convert the decision value based on the conversion function to obtain a corresponding probability value;

[0211] The fourth determination submodule is used to use the probability value as the loss probability of the high-value population.

[0212] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0213] In some optional implementations of this embodiment, the second screening module 307 includes:

[0214] The third acquisition submodule is used to obtain a high-risk loss threshold range corresponding to a preset high-value high-risk loss population;

[0215] A fourth acquisition submodule is used to acquire the first churn probability of all first agents included in the high-value population;

[0216] A second screening submodule, configured to screen out second churn probabilities that meet the high-risk churn threshold range from all the first churn probabilities;

[0217] A third screening submodule, configured to screen out second agents corresponding to the second churn probability from all the first agents;

[0218] The second generating submodule is used to generate the high-value and high-risk loss population based on the second agent.

[0219] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0220] In some optional implementations of this embodiment, the crowd recognition device based on artificial intelligence further includes:

[0221] The second acquisition module is used to acquire target information corresponding to the high-value and high-risk loss population based on a preset information type;

[0222] A second generating module, used to generate corresponding warning information based on the target information;

[0223] The third acquisition module is used to acquire the preset retention strategy content;

[0224] A determination module, used to determine a target notification method;

[0225] A sending module is used to send the target information, the warning information and the retention strategy content to relevant personnel based on the target notification method.

[0226] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the crowd recognition method based on artificial intelligence in the aforementioned implementation mode, and will not be repeated here.

[0227] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0228] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0229] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0230] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a crowd recognition method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0231] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the crowd recognition method based on artificial intelligence.

[0232] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0233] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based crowd recognition method.

[0234] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0235] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A crowd recognition method based on artificial intelligence, characterized in that: The steps include: Acquire agent portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data; Performing feature engineering processing on the portrait data to obtain corresponding first portrait data; Performing standardization processing on the first portrait data to obtain corresponding second portrait data; Performing clustering processing on the second portrait data to obtain corresponding clustering results; Screening out high-value people from all the agents based on the clustering results; Generate the churn probability of the high-value population based on a preset target classification model; Based on the loss probability, a high-value and high-risk loss population is screened out from the high-value population.

2. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: The step of performing feature engineering processing on the portrait data to obtain corresponding first portrait data specifically includes: Performing missing value filling processing on the portrait data to obtain corresponding first processed data; Encoding the first processed data to obtain corresponding second processed data; performing outlier processing on the second processed data to obtain corresponding third processed data; Performing multicollinearity elimination processing on the third processed data to obtain corresponding fourth processed data; The fourth processed data is used as the first portrait data.

3. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: The step of performing clustering processing on the second portrait data to obtain a corresponding clustering result specifically includes: Call the target clustering algorithm; Determining the number of cluster centers corresponding to the target clustering algorithm; Based on the number of cluster centers, the target clustering algorithm is used to perform clustering processing on the second portrait data to obtain the corresponding clustering result.

4. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: The step of screening out high-value people from all the agents based on the clustering result specifically includes: For each of the clustering results, obtaining the specified value attribute data of each of the clustering results; Generating a value attribute mean value of each clustering result based on the designated value attribute data of each clustering result; Sort all the clustering results by the mean values ​​of the value attributes in descending order to obtain corresponding clustering sorting results; Filter out the designated clustering result ranked first from the clustering sorting results; The designated clustering result is used as the high-value population.

5. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: The step of generating the loss probability of the high-value population based on the preset target classification model specifically includes: Acquire agent data corresponding to the high-value population; Inputting the agent data into the target classification model, and outputting the decision value corresponding to the high-value population through the target classification model; Call the preset conversion function; Convert the decision value based on the conversion function to obtain a corresponding probability value; The probability value is used as the loss probability of the high-value population.

6. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: The step of screening out a high-value and high-risk attrition population from the high-value population based on the attrition probability specifically includes: Obtain the high-risk churn threshold range corresponding to the preset high-value high-risk churn population; Obtaining the first churn probability of all first agents included in the high-value population; Filter out a second churn probability that meets the high-risk churn threshold range from all the first churn probabilities; Selecting a second agent corresponding to the second churn probability from all the first agents; The high-value and high-risk loss population is generated based on the second agent.

7. The method for crowd identification based on artificial intelligence according to claim 1, characterized in that: After the step of screening out a high-value and high-risk churn population from the high-value population based on the churn probability, the method further includes: Based on the preset information type, target information corresponding to high-value and high-risk loss groups is obtained; Generate corresponding warning information based on the target information; Obtain the preset retention strategy content; Determine the target notification method; Based on the target notification method, the target information, the warning information and the retention strategy content are sent to relevant personnel.

8. A crowd recognition device based on artificial intelligence, characterized in that: include: A first acquisition module is used to acquire the agent's portrait data; wherein the portrait data at least includes basic attribute data, value attribute data and behavior attribute data; A first processing module, used for performing feature engineering processing on the portrait data to obtain corresponding first portrait data; A second processing module, used for performing standardization processing on the first portrait data to obtain corresponding second portrait data; A clustering module, used for performing clustering processing on the second portrait data to obtain corresponding clustering results; A first screening module, used for screening out high-value people from all the agents based on the clustering result; A first generating module, used to generate the churn probability of the high-value population based on a preset target classification model; The second screening module is used to screen out high-value and high-risk churn population from the high-value population based on the churn probability.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the crowd recognition method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the crowd recognition method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images