Distribution strategy generation method and device, equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and discloses a distribution strategy generation method and device, equipment and a storage medium, and the method comprises the steps: collecting customer information of customers; according to the basic attribute data, the historical behavior data and the claim settlement complaint data, constructing corresponding features to obtain basic attribute features, historical behavior features and claim settlement complaint features; according to the basic attribute features, the historical behavior features and the claim settlement complaint features, outputting a self-service insuring willingness probability through a pre-trained insuring willingness model; pre-defining an action set and acquiring a real-time environment factor, selecting a target action from the pre-defined action set according to the self-service insuring willingness probability and the real-time environment factor, generating a corresponding distribution strategy according to the target action, and pushing the distribution strategy to the customer; the method can be applied to implementation scenes of the financial field such as insurance, the distribution strategy generation accuracy is improved, and then the customer conversion rate is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, and in particular to a generation method and device of a distribution strategy, equipment and a storage medium. BACKGROUND

[0002] Under the impetus of the digital transformation of the insurance industry, the service mode of insurance companies is undergoing unprecedented changes and challenges. With the rapid development of Internet technology and the change of customer consumption habits, the demand for self-service by customers is showing explosive growth, and self-service functions such as online insurance and policy management have become important factors for customers to choose insurance products and services.

[0003] However, there are still many problems in the service mode of current insurance companies. In terms of customer distribution strategy, traditional methods mainly rely on fixed rules and manual judgment, such as using a static threshold-based distribution strategy to guide customers to a self-service process if their self-service insurance willingness probability exceeds 0.7. However, this static strategy lacks dynamic perception of customer demand and cannot adapt to the significant differences in self-service acceptance among different customer groups. Younger customer groups usually have a high acceptance of new technology and prefer self-service processes, while older or risk-sensitive customers rely more on manual services. This single and static distribution method leads to a decrease in customer conversion rate.

[0004] It can be seen that the existing customer distribution strategy cannot dynamically adapt to changes in customer demand, resulting in a low customer conversion rate. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a generation method, device, equipment and storage medium of a distribution strategy, which mainly aims to improve the customer conversion rate through a dynamic distribution method.

[0006] In a first aspect, to solve the above technical problems, the embodiments of the present application provide a generation method of a distribution strategy, which adopts the following technical solution:

[0007] Collecting customer information of a customer, wherein the customer information includes basic attribute data, historical behavior data and claim complaint data;

[0008] According to the basic attribute data, historical behavior data and claim complaint data, corresponding features are constructed to obtain basic attribute features, historical behavior features and claim complaint features;

[0009] According to the basic attribute features, historical behavior features and claim complaint features, a pre-trained insurance willingness model outputs a self-service insurance willingness probability;

[0010] The pre-defined action set and real-time environmental factors are acquired, a target action is selected from the pre-defined action set according to the self-insurance willingness probability and the real-time environmental factors, and a corresponding distribution strategy is generated according to the target action and pushed to the customer.

[0011] In a second aspect, to solve the above technical problems, the embodiment of the present application further provides a distribution strategy generation device, which adopts the following technical scheme:

[0012] An information collection module is configured to collect customer information of the customer, wherein the customer information includes basic attribute data, historical behavior data and claim complaint data.

[0013] A feature construction module is configured to construct corresponding features according to the basic attribute data, the historical behavior data and the claim complaint data, to obtain basic attribute features, historical behavior features and claim complaint features.

[0014] A model processing module is configured to output the self-insurance willingness probability through a pre-trained insurance willingness model according to the basic attribute features, the historical behavior features and the claim complaint features.

[0015] A strategy generation module is configured to pre-define an action set and acquire real-time environmental factors, select a target action from the pre-defined action set according to the self-insurance willingness probability and the real-time environmental factors, generate a corresponding distribution strategy according to the target action and push the distribution strategy to the customer.

[0016] In a third aspect, to solve the above technical problems, the embodiment of the present application further provides a computer device, which comprises at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above distribution strategy generation method.

[0017] In a fourth aspect, to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium storing a computer program, and the computer program is executed by a processor to implement the above distribution strategy generation method.

[0018] Compared with the prior art, the embodiment of the present application has the following beneficial effects:

[0019] By pre-collecting the customer information of the customer, the customer information of the customer is used for subsequent analysis, the customer portrait can be accurately constructed, a basis is provided for subsequent insurance process distribution, and the accuracy of the distribution strategy generation is improved.

[0020] By constructing corresponding features according to the basic attribute data, the historical behavior data and the claim complaint data respectively, the basic attribute features, the historical behavior features and the claim complaint features are obtained, potential information of the data can be mined, processing of subsequent models is facilitated, probability of the model predicting the user self-insurance willingness is improved, and accuracy of the generated diversion strategy is improved;

[0021] By multi-dimensional feature analysis of the pre-trained insurance willingness model from multiple angles, customer features can be more accurately described, and the accuracy of the self-insurance willingness prediction is improved; and the model can output personalized self-insurance willingness probability according to the feature combination of the user, the accuracy of the self-insurance willingness probability prediction is improved, and the accuracy of the subsequent diversion strategy generation is improved;

[0022] By obtaining real-time environmental factors and predicting based on the environmental factors and the self-insurance willingness probability, the variables of the real-time environmental factors can be comprehensively considered, the accuracy of the target action prediction is improved, and the accuracy of the generated diversion strategy is improved. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the scheme in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0024] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0025] Figure 2 is a flowchart of one embodiment of the generation method of the diversion strategy according to the present application;

[0026] Figure 3 is a structural schematic diagram of one embodiment of the generation device of the diversion strategy according to the present application;

[0027] Figure 4 is a structural schematic diagram of one embodiment of the device according to the present application. DETAILED DESCRIPTION

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terms used in the specification are intended to describe the particular embodiments and are not intended to limit the application; the terms "include" and "have" and their any variations used in the specification and the claims and the above description of drawings are intended to cover the non-exclusive inclusion; the terms "first", "second" and the like used in the specification and the claims and the above description of drawings are intended to distinguish different objects, not to describe a particular order.

[0029] Reference herein to "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive or alternative embodiments. It is expressly understood that the embodiments described herein are combinable with each other.

[0030] In order to make the person skilled in the art better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings below.

[0031] As shown in Figure 1 The system architecture 100 can include a terminal device 101, a network 102 and a server 103, and the terminal device 101 can be a notebook computer 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103. The network 102 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0032] 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.

[0033] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing, in addition to the notebook computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an electronic book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer and a desktop computer, etc.

[0034] The server 103 can be a server providing various services, for example, a background server providing support for a page displayed on the terminal device 101.

[0035] It should be noted that the method for generating a distribution strategy provided by the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the apparatus for generating a distribution strategy is generally arranged in a server / terminal device.

[0036] It should be understood that Figure 1 The number of terminal devices, networks and servers in the system is only illustrative. According to the needs of implementation, there can be any number of terminal devices, networks and servers.

[0037] With reference to Figure 2 , a flow chart of one embodiment of the method for generating a distribution strategy according to the present application is shown. The order of steps in the flow chart can be changed according to different needs, and some steps can be omitted. The method for generating a distribution strategy provided by the embodiments of the present application can be applied to any scenario requiring generation of a distribution strategy, and then the method for generating a distribution strategy can be applied to products in these scenarios. The method for generating a distribution strategy comprises the following steps:

[0038] Step S01, collecting customer information of a customer, wherein the customer information comprises basic attribute data, historical behavior data and claim complaint data.

[0039] In the present embodiment, the electronic device (for example, the server / terminal device shown in Figure 1 ) on which the method for generating a distribution strategy runs can acquire the customer information through a wired connection mode or a wireless connection mode. It should be noted that the wireless connection mode can include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other now known or future developed wireless connection modes.

[0040] In this embodiment, customer information is collected, including basic attribute information, historical behavior data, and claims complaint data. Basic customer information includes the customer's age, gender, occupation, etc. Historical behavior data refers to user behavior data on pages such as the APP, such as the frequency of clicking the "Insure Now" button, form completion rate, number of price inquiries in the past 3 months, number of consecutive years of coverage, total number / amount of insurance products purchased historically, number of valid policies, and historical self-service insurance application rate, etc. Claims complaint data includes the number of complaints, reasons for complaints (such as "poor service attitude," "complex claims process," etc.), causes of accidents (such as "natural disasters," "human-caused accidents," "health problems," etc.), and claims timeliness rating, etc.

[0041] In this embodiment, the behavior database, customer management database, claims database, and complaint database are invoked. Historical behavior data, basic attribute data, claims data, and complaint data are extracted from each database using specific query statements (such as SQL query statements). The claims data and complaint data are aggregated to obtain claims and complaint data. The extracted data is then preprocessed, including data cleaning, such as removing duplicate data and correcting data with format errors. The data is then transformed to unify data from different sources into a format that is easy to store and analyze (such as unifying the date format to YYYY-MM-DD).

[0042] In this embodiment, by collecting customer information in advance and using it for subsequent analysis, a customer profile can be accurately constructed, providing a basis for subsequent insurance process diversion and improving the accuracy of diversion strategy generation.

[0043] In one embodiment, the method further includes, prior to collecting the customer's customer information:

[0044] At least one interactive page should be pre-built according to business needs, and event tracking should be embedded in each interactive page.

[0045] Real-time monitoring of behavioral data on each interactive page based on data tracking points;

[0046] Behavioral data is collected in real time or periodically by embedding data points, and the behavioral data is encapsulated according to a predefined data format to obtain historical behavioral data, which is then stored in a preset behavioral database.

[0047] In this embodiment, at least one interactive page (such as an insurance product display page or an insurance application form page) is pre-built in the insurance application APP or mini-program according to business needs (such as insurance application needs). Tracking points are set on each interactive page. Tracking points include, but are not limited to, front-end tracking, back-end tracking, visual tracking, and no-code tracking (full tracking). Front-end tracking involves directly inserting tracking code into the client-side code. When a user action triggers the tracking point, the front-end tracking point sends the collected behavioral data to the database. Back-end tracking involves setting tracking points in the back-end service code. When a user action involves a back-end interface call, the back-end code records the relevant data. Visual tracking refers to using visualization tools to track data on interactive pages. On the page, you can directly select the elements you want to monitor without writing any code. No-code (full-code) monitoring refers to pre-integrating a general data collection SDK. This SDK automatically collects all user action data within the application, and then filters and analyzes it as needed during the data analysis phase. Selecting appropriate tracking methods to embed tracking points on interactive pages allows for monitoring of user action data on each interactive page. The tracking points also collect user behavior data in real time on each interactive page, encapsulate this behavior data according to a predetermined format to obtain historical behavior data, and store this historical behavior data in a preset behavior database for subsequent user queries.

[0048] In this embodiment, by embedding appropriate tracking points when constructing each interactive page, customer behavior data can be accurately captured, thereby improving the accuracy of subsequent traffic diversion strategy generation. Furthermore, monitoring and collecting data based on tracking points can promptly obtain the latest user operation information, achieving real-time data collection. This allows for the generation of subsequent traffic diversion strategies based on the latest data, improving the accuracy of traffic diversion strategy generation.

[0049] Step S202: Construct corresponding features based on basic attribute data, historical behavior data, and claims complaint data to obtain basic attribute features, historical behavior features, and claims complaint features;

[0050] In this embodiment, basic attribute features are constructed based on basic attribute data, historical behavior features are constructed based on historical behavior data, and claims complaint features are constructed based on claims complaint data.

[0051] In this embodiment, according to the basic attribute data obtained in step S201, an encoding operation is performed based on each item of the basic attribute data, and the encoding of each item is collected to obtain the basic attribute feature. The construction of the basic attribute feature from the basic attribute data specifically includes dividing the age into different intervals, and encoding based on the age interval, such as 0-18 years old (encoded as 1), 19-30 years old (encoded as 2), 31-50 years old (encoded as 3), and 51 years old and above (encoded as 4); encoding based on gender, such as male (1) and female (0); encoding based on occupation, such as dividing the occupation into white-collar (encoded as 1), blue-collar (encoded as 2), freelancer (encoded as 3), and student (encoded as 4); encoding based on region, such as dividing the region where the customer is located into first-tier cities (encoded as 1), second-tier cities (encoded as 2), and third-tier cities and below (encoded as 3); and encoding based on device type, such as dividing the device type into mobile terminal (encoded as 1) and PC terminal (encoded as 0). The needs of different age groups and genders for self-service are different, and the above are encoded to construct the basic attribute feature.

[0052] In this embodiment, the historical behavior feature is constructed according to the historical behavior data. Specifically, based on the statistical behavior, the number of times that the customer clicks the “immediate insurance” button within a certain period of time is counted and normalized; based on the form completion rate, the completion rate of the customer filling out the insurance form is calculated, that is, the proportion of the number of completed forms to the total number of accessed forms, and normalized; based on the number of inquiries, the number of times that the customer makes insurance inquiries in the past three months is counted and normalized; and the like. The features after the above normalization are collected to obtain the historical behavior feature.

[0053] In this embodiment, the claim complaint feature is constructed according to the claim complaint data. Specifically, based on the total number of complaints, the total number of complaints of the customer within a certain period of time is counted and normalized; based on the complaint reason, the complaint reason is divided into poor service attitude (encoded as 1), complex claim process (encoded as 2), and unclear terms (encoded as 3); based on the cause of the accident, the cause of the accident is divided into natural disaster (encoded as 1), man-made accident (encoded as 2), and health problem (encoded as 3); and based on the claim timeliness score, the claim timeliness score is calculated according to the customer's evaluation of the claim timeliness (such as 1-5 points), and normalized.

[0054] In one embodiment, the corresponding features are constructed according to the basic attribute data, the historical behavior data, and the claim complaint data, respectively, to obtain the basic attribute feature, the historical behavior feature, and the claim complaint feature. The steps include:

[0055] The data types of each data in the basic attribute data are identified, the basic attribute data is encoded based on the data types, and the basic attribute feature is obtained.

[0056] The historical behavior data is intercepted based on a time sequence to obtain historical behavior data of multiple time periods, a ratio of each data in the historical behavior data of each time period is calculated, feature extraction is performed based on the ratio to obtain historical behavior features;

[0057] Keywords of the claim complaint data are extracted, a claim complaint category is identified based on the keywords, the claim complaint category is encoded to obtain claim complaint features.

[0058] In this embodiment, the basic attribute data is identified item by item, the attribute of each data item is identified, and the basic attribute data is divided into different data types according to the attribute of each data item, wherein the data types include a numerical type and a discrete type, a preset encoding method is used for encoding based on the discrete type, the preset encoding method includes label encoding and one-hot encoding, for the numerical type, a normalization method or a standardization method is used for processing, and all processed data is collected to obtain basic attribute features; for the historical behavior data, a suitable time granularity is determined to intercept the historical behavior data according to business requirements and analysis purposes, the time granularity can be day, week, month, etc., for example, the behavior trend of a customer in a long time can be analyzed by selecting data interception per month; the behavior trend of a customer in a short time can be analyzed by selecting data interception per day, the historical behavior data is divided into multiple time periods according to the determined time granularity, the ratio of each data is calculated for the historical behavior data in each time period, wherein the ratio includes the claim frequency in a preset time period, the form filling completion rate, the quotation ratio, etc., feature extraction is performed based on each ratio, and the features corresponding to each data are collected to obtain historical behavior features; for the claim complaint data, the claim complaint data is generally in the form of text, after obtaining the claim complaint data, a pretreatment operation is performed on the claim complaint data to remove irrelevant characters and punctuation marks, and a word segmentation processing is performed to obtain a word segmentation set, keywords are identified based on a TF-IDF algorithm to obtain the keywords, different claim complaint categories are defined, the keywords are mapped to different claim complaint categories, according to the specific content of different categories (such as poor service attitude, etc.), encoding is performed by using one-hot encoding to obtain the encoding corresponding to the specific content of each category, and all encodings are collected to obtain claim complaint features.

[0059] In this embodiment, by constructing corresponding features based on the basic attribute data, the historical behavior data and the claim complaint data respectively, the basic attribute features, the historical behavior features and the claim complaint features are obtained, the potential information of the data can be mined, the subsequent model processing is facilitated, the probability of the model predicting the user self-service claim willingness is improved, and the accuracy of the diversion strategy generation is improved.

[0060] In step S203, the self-insurance willingness probability is output by the pre-trained insurance willingness model according to the basic attribute features, the historical behavior features and the claim complaint features.

[0061] In the embodiment, the pre-trained insurance willingness model is a machine learning algorithm based on Gradient Boosted Decision Trees (GBDT), and preferably, the LightGBM (Light Gradient Boosting Machine, LGBM) is used to construct the insurance willingness model.

[0062] In the embodiment, the basic attribute features, the historical behavior features and the claim complaint features constructed in step S202 are input into the pre-trained insurance willingness model, and the insurance willingness model is used to process the features to output the self-insurance willingness probability.

[0063] In the embodiment, the pre-trained insurance willingness model is used to analyze the multi-dimensional features from multiple angles, which can more accurately describe the customer features and improve the accuracy of the self-insurance willingness prediction. Moreover, the model can output the personalized self-insurance willingness probability according to the feature combination of the user, thereby improving the accuracy of the self-insurance willingness probability prediction and the accuracy of the subsequent shunt strategy generation.

[0064] In one embodiment, the step of outputting the self-insurance willingness probability by the pre-trained insurance willingness model according to the basic attribute features, the historical behavior features and the claim complaint features includes:

[0065] The basic attribute features, the historical behavior features and the claim complaint features are integrated to obtain a customer feature set;

[0066] The customer feature set is input into the pre-trained insurance willingness model, and a histogram algorithm is used to discretize each feature in the customer feature set to obtain a discrete interval corresponding to each feature;

[0067] Each discrete interval corresponding to each feature is input into a corresponding decision tree, and each discrete interval corresponding to each feature is assigned to a corresponding leaf node according to a pre-set classification rule to obtain a plurality of leaf node values;

[0068] The plurality of leaf node values are weighted and summed to obtain the self-insurance willingness probability.

[0069] In this embodiment, the basic attribute features, historical behavior features and claim complaint features obtained in step S202 are integrated into a customer feature set, a pre-trained insurance willingness model is loaded, the customer feature set is input into the pre-trained insurance willingness model, the model uses a histogram algorithm to first discretize each feature in the customer feature set, since the above has been encoded according to the feature interval, at this time, only the discrete interval of the feature needs to be identified, the discrete interval corresponding to each feature is obtained, and the discretized features are sequentially input into the decision tree of the insurance willingness model. Starting from the root node of the decision tree, according to the preset classification rule, enter the leaf node corresponding to the decision tree, the classification rule refers to assigning the corresponding leaf node according to the belonging discrete interval, for example, “age less than or equal to 30 years old”, the samples in the 0-30 age interval will enter the left child node, and the samples greater than 30 years old will enter the right child node; based on the leaf node of each feature, the class distribution of all features in the leaf node is calculated, for example, there are 10 positive class samples and 5 negative class samples in a certain leaf node, and the probability of the positive class output by the leaf node is 10 / (10+5)=0.66; by processing the corresponding features through all decision trees, a plurality of leaf node values are obtained, the plurality of leaf node values are weighted and summed based on the weight of each decision tree, and finally the self-service insurance willingness probability is obtained.

[0070] In this embodiment, the multi-dimensional information of the customer is processed by the insurance willingness model, the histogram algorithm is used to discretize each feature, which greatly reduces the calculation amount and improves the prediction speed of the insurance willingness model; the discretized features are input into the decision tree, the features are assigned to the corresponding leaf nodes according to the pre-set classification rule, the leaf node values are obtained, and the self-service insurance willingness probability is obtained by weighted sum of the plurality of leaf node values, which comprehensively considers the prediction results of the plurality of decision trees, can more accurately reflect the possibility of customer self-service insurance, and improves the accuracy of the self-service insurance willingness probability.

[0071] Step S204, pre-defining an action set and obtaining real-time environmental factors, selecting a target action from the pre-defined action set according to the self-service insurance willingness probability and the real-time environmental factors, generating a corresponding shunt strategy according to the target action and pushing it to the customer.

[0072] In the embodiment, the predefined action set can include A1 representing guiding a self-service process, A2 representing transferring to a human agent, and A3 representing a hybrid model (self-service + key node human assistance), A1 includes directly guiding the customer to a self-service insurance application page, the page provides detailed insurance application instructions, product information display, and a convenient insurance application operation portal; A2 includes immediately transferring the customer to a human customer service, communicating with the customer one-on-one through the human customer service, understanding the customer's needs and providing personalized insurance application suggestions; A3 includes that the customer first performs preliminary operations on the self-service insurance application page, and when reaching a key node (such as product selection, health notification problem, etc.), an automatic pop-up human customer service dialog box or a one-key transfer button of human customer service is provided, and the human customer service is intervened to provide assistance; the above implementation environment factors include but are not limited to accurately recording the access time of the customer, distinguishing between working hours and non-working hours; identifying the type of device used by the customer by identifying the request information sent by the customer, such as a mobile phone, a computer, etc.; the predefined action set is obtained according to the real-time environment factors of the customer, the real-time environment factors and the self-service insurance application willingness probability are input into the pre-trained action selection model, the target willingness action probability is predicted through the model, the target action is selected from the predefined action set based on the target willingness action probability, and the corresponding distribution strategy is generated based on the target action and pushed to the customer, wherein the distribution strategy is generated according to the target action of the customer (for example, the target action of the customer is to guide the self-service process, then the customer is guided to the self-service insurance application page, the page provides detailed insurance application instructions, product information display, and a convenient insurance application operation portal); the pre-trained action selection model refers to a deep neural network CNN, and the model includes an input layer, multiple hidden layers, and an output layer.

[0073] In the embodiment, by obtaining the real-time environment factors and predicting based on the environment factors and the self-service insurance application willingness probability, the variables of the real-time environment factors can be comprehensively considered, the accuracy of the target action prediction is improved, and the accuracy of the distribution strategy generation is improved.

[0074] In one embodiment, the target action is selected from the predefined action set according to the self-service insurance application willingness probability and the real-time environment factors, including:

[0075] The real-time environment factors are encoded to obtain real-time environment factor encoding;

[0076] The real-time environment factor encoding and the self-service insurance application willingness probability are combined to obtain a comprehensive feature vector;

[0077] The pre-trained action selection model is used to process the comprehensive feature vector to generate a target willingness action probability;

[0078] The target action is selected from the predefined action set according to the target willingness action probability.

[0079] In this embodiment, the real-time environmental factors are encoded by using a preset encoding method to obtain real-time environmental factor encoding, the preset encoding method includes but is not limited to One-Hot Encoding, etc. The encoded real-time environmental factors are spliced with the self-insurance willingness probability to form a comprehensive feature vector. The comprehensive feature vector is input into a pre-trained action selection model. The comprehensive feature vector is processed by the action selection model to output a target willingness action probability. According to the target willingness action probability, it is determined which interval of the pre-defined action set the target willingness action probability belongs to. The target action is determined from the pre-defined action set according to the interval.

[0080] In this embodiment, the real-time environmental factors are encoded by using a preset encoding method to obtain real-time environmental factor encoding, the preset encoding method includes but is not limited to One-Hot Encoding, etc. The encoded real-time environmental factors are spliced with the self-insurance willingness probability to form a comprehensive feature vector. The comprehensive feature vector is input into a pre-trained action selection model. The comprehensive feature vector is processed by the action selection model to output a target willingness action probability. According to the target willingness action probability, it is determined which interval of the pre-defined action set the target willingness action probability belongs to. The target action is determined from the pre-defined action set according to the interval.

[0081] In another embodiment, the step of processing the comprehensive feature vector by using the pre-trained action selection model to generate the target willingness action probability includes:

[0082] The comprehensive feature vector is input through the input layer of the action selection model.

[0083] The neurons in each hidden layer of the action selection model are used to weight and sum the comprehensive feature vector, and the target feature vector is obtained by linear transformation through a first activation function.

[0084] The target feature vector is converted into a probability by a second activation function to obtain the target willingness action probability.

[0085] In this embodiment, the comprehensive feature vector is input through the input layer of the action selection model. The input layer receives and transmits the comprehensive feature vector to multiple hidden layers of the action selection model for processing. In each hidden layer of the model, the neurons weight and sum the input features. The input of each neuron is obtained by multiplying the output of the previous layer with the corresponding weight and adding a bias term. For each layer, the output can be represented as z = w * x + b, where w is the weight matrix, x is the input vector, and b is the bias vector. The weighted sum result is nonlinearly transformed by a first activation function (such as ReLU). After processing by the hidden layers, the target feature vector is obtained. The target feature vector is input into the output layer, and the target willingness action probability is obtained by probability conversion through a second activation function (such as Softmax).

[0086] In the embodiment, the complex relationship between the input features can be captured by processing the spliced comprehensive feature vector through each hidden layer of the target will action model, and the target will action probability can be obtained by performing probability conversion on the target feature vector through the second activation function, thereby improving the accuracy of the target will action probability generation, and further improving the accuracy of the subsequent shunt strategy generation.

[0087] In one embodiment, after the corresponding shunt strategy is generated according to the target action and pushed to the customer, the method further comprises:

[0088] S501, creating a Q table and predefining a reward function;

[0089] S502, in the Q table, selecting an action in the current state through an epsilon-greedy strategy, executing the action, and calculating the reward value of the action according to the reward function;

[0090] S503, updating the Q table according to the current state, the action, the reward value, and the maximum Q value of the next state through the updating rule of Q-Learning;

[0091] Steps S502-S503 are repeated until the iteration number reaches the preset iteration number, the calculation is stopped to obtain a calculation result, and the action strategy corresponding to the current state is strengthened according to the calculation result.

[0092] In this embodiment, a Q table is created, the Q table is a two-dimensional array for recording the expected reward value of each state-action, each element Q[s][a] represents the expected reward of taking action a in state s, a reward function is predefined, the reward function includes customer conversion rate and service cost, wherein the customer conversion rate refers to the number of customers who successfully self-insure / total number of customers, and the service cost refers to resource consumption of manual service, such as agent call duration, etc., the formula of the reward function includes r = a x customer conversion rate - b x service cost, wherein r refers to the reward function, a and b refer to weights; an e-greedy strategy is adopted to select the action with the maximum Q value in the current state and execute, the action refers to the above predefined action set, the selected action is executed, the customer conversion rate and the service cost are observed, the reward value of the action in the current state is calculated using the reward function according to the customer conversion rate and the service cost, and the Q table is updated according to the current state, the action, the reward value and the maximum Q value of the next state through the updating rule of Q-Learning, the formula of the updating rule of Q-Learning includes: Q[s][a] = Q[s][a] + a x (r + g x max(Q[s'][a']) - Q[s][a]), wherein Q[s][a] represents the current expected reward value (i.e. Q value) of taking action a in state s, a refers to the learning rate; r refers to the reward value, g refers to the discount factor; max(Q[s'][a']) refers to the maximum Q value in all possible actions a' in the next state s'; steps S502-S503 are repeated until the iteration number reaches the preset iteration number, the calculation is stopped to obtain a calculation result, and the action strategy corresponding to the current state is strengthened according to the calculation result.

[0093] In this embodiment, by defining a reward function, initializing a Q table, and selecting states and actions and updating Q values according to the Q-Learning algorithm, the model can find a balance point between customer conversion rate and cost control, thereby improving the accuracy of subsequent shunt strategy generation.

[0094] It should be emphasized that, in order to further ensure the privacy and security of the above-mentioned obtaining customer information, etc., the above-mentioned obtaining customer information, etc. can also be stored in a node of a blockchain.

[0095] The blockchain referred to in the present application is a new application mode of distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm and other computer technologies. Blockchain, in essence, is a decentralized database, a series of data blocks associated using cryptographic methods, each containing information about a batch of network transactions, used to verify the validity of the information (anti-fake) and generate the next block. The blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer, etc.

[0096] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. The artificial intelligence (AI) is a theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by a digital computer, perceiving an environment, acquiring knowledge and using the knowledge to obtain optimal results.

[0097] The artificial intelligence basic technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0098] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related 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 above-mentioned embodiments of each method. Among them, the storage medium can be a non-volatile storage medium such as a magnetic disc, an optical disc, a read-only memory (ROM), or a random access memory (RAM).

[0099] It should be understood that, although each step in the flowchart of the accompanying drawings is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise stated herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other orders. Moreover, at least part of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0100] Further referring to Figure 3 , as an implementation of the method shown in Figure 2 , the present application provides an embodiment of a device for generating a shunt strategy. The device embodiment corresponds to the method embodiment shown in Figure 2 , and the device can be specifically applied to various computer devices.

[0101] As Figure 3As shown, the generation device 300 of the shunt strategy in the embodiment includes an information collection module 301, a feature construction module 302, a model processing module 303, and a strategy generation module 304. Among them:

[0102] The information collection module 301 is configured to collect customer information of the customer, wherein the customer information includes basic attribute data, historical behavior data, and claim complaint data.

[0103] In one embodiment, the information collection module includes:

[0104] The point implantation sub-module is configured to pre-construct at least one interactive page according to business requirements, and implant a point in each interactive page.

[0105] The data monitoring sub-module is configured to monitor behavior data on each interactive page in real time based on the point.

[0106] The data packaging sub-module is configured to collect behavior data in real time or periodically through the point, package the behavior data according to a predefined data format, obtain historical behavior data, and store the historical behavior data into a preset behavior database.

[0107] The feature construction module 302 is configured to construct corresponding features according to the basic attribute data, the historical behavior data, and the claim complaint data, respectively, to obtain basic attribute features, historical behavior features, and claim complaint features.

[0108] In one embodiment, the feature construction module includes:

[0109] The first feature construction sub-module is configured to identify data types of each data in the basic attribute data, encode the basic attribute data based on the data types, and obtain the basic attribute features.

[0110] The second feature construction sub-module is configured to intercept the historical behavior data based on a time sequence, obtain historical behavior data of multiple time periods, calculate ratios of each data in the historical behavior data of each time period, perform feature extraction based on the ratios, and obtain the historical behavior features.

[0111] The third feature construction sub-module is configured to extract keywords of the claim complaint data, identify claim complaint categories based on the keywords, encode the claim complaint categories, and obtain the claim complaint features.

[0112] The model processing module 303 is configured to output a self-insurance willingness probability through a pre-trained insurance willingness model according to the basic attribute features, the historical behavior features, and the claim complaint features.

[0113] In one embodiment, the model processing module includes:

[0114] The feature integration module is configured to integrate the basic attribute features, the historical behavior features, and the claim complaint features to obtain a customer feature set;

[0115] The discrete processing submodule is configured to input the customer feature set into a pre-trained insurance application willingness model, and perform discrete processing on each feature in the customer feature set by using a histogram algorithm to obtain a discrete interval corresponding to each feature;

[0116] The leaf node value calculation submodule is configured to input the discrete interval corresponding to each feature into a corresponding decision tree, and assign the discrete interval corresponding to each feature to a corresponding leaf node according to a pre-set classification rule to obtain a plurality of leaf node values.

[0117] The leaf node value summation submodule is configured to perform weighted summation on the plurality of leaf node values to obtain the self-service insurance application willingness probability.

[0118] The strategy generation module 304 is configured to predefine an action set, obtain real-time environmental factors, select a target action from the predefined action set according to the self-service insurance application willingness probability and the real-time environmental factors, and generate a corresponding shunt strategy according to the target action and push the shunt strategy to the customer.

[0119] In one embodiment, the strategy generation module includes:

[0120] The encoding submodule is configured to perform an encoding operation on the real-time environmental factors to obtain real-time environmental factor encodings.

[0121] The concatenation submodule is configured to combine the real-time environmental factor encodings and the self-service insurance application willingness probability to obtain a comprehensive feature vector.

[0122] The probability generation submodule is configured to process the comprehensive feature vector by using a pre-trained action selection model to generate a target willingness action probability.

[0123] The action screening submodule is configured to screen a target action from the predefined action set according to the target willingness action probability.

[0124] In another embodiment, the probability generation submodule includes:

[0125] The model input subunit is configured to input the comprehensive feature vector through an input layer of the action selection model.

[0126] The model processing subunit is configured to perform weighted summation on the comprehensive feature vector by using neurons in each hidden layer of the action selection model, and perform linear transformation on the comprehensive feature vector by using a first activation function to obtain a target feature vector.

[0127] The model conversion subunit is configured to perform probability conversion on the target feature vector by using a second activation function to obtain the target willingness action probability.

[0128] In one embodiment, the device further comprises:

[0129] a creating module configured to create a Q table and a predefined reward function, S501;

[0130] a reward value calculating module configured to select an action in the current state in the Q table by an e-greedy strategy, execute the action, and calculate a reward value of the action according to the reward function, S502;

[0131] a Q table updating module configured to update the Q table according to the current state, the action, the reward value, and a maximum Q value of the next state by an updating rule of Q-Learning, S503;

[0132] an iteration module configured to repeat steps S502-S503 until a preset iteration number is reached, stop calculation to obtain a calculation result, and strengthen the action strategy corresponding to the current state according to the calculation result.

[0133] In this embodiment, the customer information of the customer is collected in advance, and the customer information of the customer is used for subsequent analysis, so that the customer portrait can be accurately constructed, and a basis is provided for subsequent insurance process distribution, and the accuracy of the generated distribution strategy is improved.

[0134] The basic attribute feature, the historical behavior feature, and the claim complaint feature are obtained by constructing corresponding features according to the basic attribute data, the historical behavior data, and the claim complaint data, potential information of the data can be mined, subsequent model processing is facilitated, the probability of predicting the self-service insurance willingness of the user by the model is improved, and the accuracy of the generated distribution strategy is improved.

[0135] The customer features can be more accurately described by performing multi-angle analysis on the multi-dimensional features of the pre-trained insurance willingness model, and the accuracy of the self-service insurance willingness prediction is improved. The model can output personalized self-service insurance willingness probability according to the feature combination of the user, the accuracy of the self-service insurance willingness probability prediction is improved, and the accuracy of the subsequent distribution strategy generation is improved.

[0136] The real-time environmental factors are obtained, and the self-service insurance willingness probability is predicted based on the environmental factors, so that the variables of the real-time environmental factors can be comprehensively considered, the accuracy of the target action prediction is improved, and the accuracy of the distribution strategy generation is improved.

[0137] To solve the above technical problems, the embodiment of the present application also provides a device (computer device). For details, please refer to Figure 4 , Figure 4 The basic structure block diagram of the computer device of the embodiment is shown in the figure.

[0138] The computer device 4 includes a memory 41, a processor 42, and a network interface 43, which are communicatively connected by a system bus. It should be noted that the computer device 4 is only shown with the memory 41, the processor 42, and the network interface 43, but it should be understood that not all of the illustrated components are required to be implemented, and more or fewer components can be alternatively implemented. Among them, those skilled in the art can understand that the computer device herein is a device capable of automatically performing numerical calculation 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 (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0139] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The computer device can interact with the user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, and the like.

[0140] The memory 41 includes at least one type of readable storage medium, which includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or a 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, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Of course, the memory 41 can also include both the internal storage unit and the external storage device of the computer device 4. In the present embodiment, the memory 41 is generally used to store an operating system and various application software installed in the computer device 4, such as computer readable instructions of the flow splitting method, and the like. In addition, the memory 41 can also be used to temporarily store various data that have been output or will be output.

[0141] The processor 42 may, in some embodiments, be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In the present embodiment, the processor 42 is configured to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions of the generation method of the distribution strategy.

[0142] The network interface 43 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0143] In the implementation of the electronic device of the present application, the customer information of the customer is collected in advance, and the customer information of the customer is used for subsequent analysis, so that the customer portrait can be accurately constructed, which provides a basis for subsequent insurance process distribution, and improves the accuracy of the generation of the distribution strategy.

[0144] By constructing corresponding features according to the basic attribute data, the historical behavior data and the claim complaint data, the basic attribute features, the historical behavior features and the claim complaint features are obtained, the potential information of the data can be mined, the subsequent model processing is facilitated, the probability of predicting the self-insurance willingness of the user is improved, and the accuracy of the generation of the distribution strategy is improved.

[0145] By performing multi-angle analysis on the multi-dimensional features of the pre-trained insurance willingness model, the customer features can be more accurately described, and the accuracy of the self-insurance willingness prediction is improved. Moreover, the model can output an individualized self-insurance willingness probability according to the feature combination of the user, the accuracy of the self-insurance willingness probability prediction is improved, and the accuracy of the generation of the subsequent distribution strategy is improved.

[0146] By obtaining real-time environmental factors and predicting based on the environmental factors and the self-insurance willingness probability, the variables of the real-time environmental factors can be comprehensively considered, the accuracy of the target action prediction is improved, and the accuracy of the generation of the distribution strategy is improved.

[0147] The present application also provides another implementation, that is, to provide a storage medium (computer readable storage medium), the computer readable storage medium stores computer readable instructions, the computer readable instructions can be executed by at least one processor, so that the at least one processor executes the steps of the generation method of the distribution strategy as described above.

[0148] In the implementation of the computer readable storage medium of the present application, the customer information of the customer is collected in advance, and the customer information of the customer is used for subsequent analysis, so that the customer portrait can be accurately constructed, which provides a basis for subsequent insurance process distribution, and improves the accuracy of the generation of the distribution strategy.

[0149] By constructing corresponding features according to the basic attribute data, the historical behavior data and the claim complaint data respectively, the basic attribute features, the historical behavior features and the claim complaint features are obtained, potential information of the data can be mined, subsequent model processing is facilitated, the probability of predicting the user self-insurance willingness is improved, and the accuracy of the generated diversion strategy is improved;

[0150] Through multi-dimensional feature analysis of the pre-trained insurance willingness model from multiple angles, the customer features can be more accurately described, and the accuracy of the self-insurance willingness prediction is improved; and the model can output personalized self-insurance willingness probability according to the feature combination of the user, the accuracy of the self-insurance willingness probability prediction is improved, and the accuracy of the subsequent diversion strategy generation is improved;

[0151] By obtaining real-time environmental factors and predicting based on the environmental factors and the self-insurance willingness probability, the variables of the real-time environmental factors can be comprehensively considered, the accuracy of the target action prediction is improved, and the accuracy of the generated diversion strategy is improved.

[0152] The non-company software tools or components appearing in the embodiments of the present application are only illustrative and do not represent actual use.

[0153] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and a general hardware platform, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application 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 plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the methods of various embodiments of the present application.

[0154] Obviously, the above-described embodiments are only some of the embodiments of the present application, not all the embodiments, and the preferred embodiments of the present application are given in the drawings, but do not limit the patent scope of the present application. The present application can be realized in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacements to some technical features. Any equivalent structure made by using the contents of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the scope of the patent protection of the present application.

Claims

1. A method for generating a split strategy, characterized in that, The methods include: Collect customer information, which includes basic attribute data, historical behavior data, and claims and complaint data; Based on the basic attribute data, historical behavior data, and claims complaint data, corresponding features are constructed to obtain the basic attribute features, historical behavior features, and claims complaint features. Based on basic attribute characteristics, historical behavior characteristics, and claims complaint characteristics, the probability of self-service insurance purchase intention is output through a pre-trained insurance purchase intention model. Predefine a set of actions and obtain real-time environmental factors. Based on the probability of self-service insurance purchase intention and real-time environmental factors, select target actions from the predefined set of actions, generate corresponding triage strategies based on the target actions, and push them to customers.

2. The method of generating a split strategy of claim 1, wherein, Before collecting customer information, the method also includes: At least one interactive page should be pre-built according to business needs, and event tracking should be embedded in each interactive page. Real-time monitoring of behavioral data on each interactive page based on data tracking points; Behavioral data is collected in real time or periodically by embedding data points, and the behavioral data is encapsulated according to a predefined data format to obtain historical behavioral data, which is then stored in a preset behavioral database.

3. The method of generating a split strategy of claim 1, wherein, The steps to construct corresponding features based on basic attribute data, historical behavior data, and claims complaint data, and to obtain basic attribute features, historical behavior features, and claims complaint features include: Identify the data type of each data in the basic attribute data, encode the basic attribute data based on the data type, and obtain the basic attribute features; Historical behavior data is extracted based on time series data to obtain historical behavior data for multiple time periods. The ratio of each data point in the historical behavior data for each time period is calculated, and feature extraction is performed based on the ratio to obtain historical behavior features. Extract keywords from claims complaint data, identify claim complaint categories based on keywords, encode claim complaint categories, and obtain claims complaint characteristics.

4. The method of generating a split strategy of claim 1, wherein, Based on basic attribute characteristics, historical behavioral characteristics, and claims complaint characteristics, the steps to output the probability of self-service insurance purchase intention using a pre-trained insurance purchase intention model include: By integrating basic attribute features, historical behavior features, and claims and complaint features, a customer feature set is obtained; The customer feature set is input into the pre-trained insurance intention model, and the histogram algorithm is used to discretize each feature in the customer feature set to obtain the discrete interval corresponding to each feature. The discrete interval corresponding to each feature is input into the corresponding decision tree. According to the pre-set classification rules, the discrete interval corresponding to each feature is assigned to the corresponding leaf node, resulting in multiple leaf node values. The probability of self-service insurance purchase intention is obtained by weighted summation of multiple leaf node values.

5. The method of generating a split strategy of claim 1, wherein, Based on the probability of self-service insurance purchase intention and real-time environmental factors, target actions are selected from a predefined action set, including: The real-time environmental factors are encoded to obtain the real-time environmental factor codes. By combining real-time environmental factor encoding and the probability of self-service insurance purchase intention, a comprehensive feature vector is obtained; The comprehensive feature vector is processed using a pre-trained action selection model to generate the probability of the target's intended action; Based on the probability of the desired action, the target action is selected from a predefined set of actions.

6. The method of generating a split strategy of claim 5, wherein, The step of processing the comprehensive feature vector by using the pre-trained action selection model to generate the target willingness action probability comprises: inputting the comprehensive feature vector through an input layer of the action selection model; performing weighted summation on the comprehensive feature vector by using neurons in each hidden layer of the action selection model, and performing linear transformation on the weighted summation by using a first activation function to obtain a target feature vector; performing probability conversion on the target feature vector by using a second activation function to obtain the target willingness action probability.

7. The method of generating a split strategy of claim 1, wherein, After generating the corresponding distribution strategy according to the target action and pushing the distribution strategy to the customer, the method further comprises: S501, creating a Q table and a pre-defined reward function; S502, in the Q table, selecting an action in the current state by using an ε-greedy strategy, executing the action, and calculating a reward value of the action according to the reward function; S503, updating the Q table according to the current state, the action, the reward value, and the maximum Q value of the next state by using an updating rule of Q-Learning; repeating steps S502-S503 until the iteration number reaches a pre-set iteration number, stopping calculation to obtain a calculation result, and strengthening the action strategy corresponding to the current state according to the calculation result.

8. An apparatus for generating a split strategy, the apparatus comprising: The device comprises: an information collection module configured to collect customer information of a customer, wherein the customer information comprises basic attribute data, historical behavior data, and claim complaint data; a feature construction module configured to construct corresponding features from the basic attribute data, the historical behavior data, and the claim complaint data to obtain basic attribute features, historical behavior features, and claim complaint features; a model processing module configured to output a self-service insurance willingness probability by using a pre-trained insurance willingness model according to the basic attribute features, the historical behavior features, and the claim complaint features; a strategy generation module configured to pre-define an action set, acquire real-time environmental factors, select a target action from the pre-defined action set according to the self-service insurance willingness probability and the real-time environmental factors, and generate a corresponding distribution strategy according to the target action and push the distribution strategy to the customer.

9. A computer device, comprising: The computer device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for generating a distribution strategy according to any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the method for generating a distribution strategy according to any one of claims 1 to 7.

Images