Method for efficiently pushing business opportunities based on feature selection

Abstract

The invention discloses an efficient business opportunity pushing method based on feature selection, and relates to the technical field of big data analysis, historical data of clients and data collected by business personnel in client contact are integrated, and the integrated data are preprocessed; extracting features related to insurance purchase willingness from the integrated customer data set, and encoding the extracted features to obtain a basic feature sequence; by collecting and analyzing the multi-dimensional data of the customers, constructing the accurate customer portraits and based on accurate matching of the customer portraits, the pushing accuracy is greatly improved, invalid pushing is reduced, the satisfaction degree of the customers to pushing is improved, and due to the fact that the pushed content is highly matched with the customer requirements, the customers are easier to generate purchase willingness, and the customer experience is improved. In addition, the investment of invalid customers can be reduced through accurate pushing, more resources are input to potential high-value customers, and optimal configuration of the resources is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a method for efficiently pushing business opportunities based on feature selection. Background Art

[0002] With the continuous growth of the economy and the improvement of people's risk awareness, the insurance industry has ushered in unprecedented development opportunities. Insurance companies continue to launch new products and services to meet customers' growing insurance needs. At the same time, technological advances have also provided strong technical support for the development of the insurance industry, enabling insurance companies to handle business more efficiently and improve service quality. Against the background of the rapid development of the insurance industry, market competition has become increasingly fierce. Many insurance companies have stepped up their marketing efforts to compete for customer resources. The feature-based push method is very critical in digital marketing, mobile applications and personalized services. This method customizes and pushes personalized content by analyzing user behavior data, preferences, contextual information and other relevant features.

[0003] The existing feature selection methods lack verification, and the determined features lack weights, which cannot guarantee the push effect. In addition, traditional marketing methods are often inefficient and difficult to accurately locate customer portraits with high purchasing intentions in the insurance industry. Therefore, how to conduct accurate customer portraits and assist business personnel in precise marketing and services is the problem we need to solve. To this end, a method for efficiently pushing business opportunities based on feature selection is proposed. Summary of the invention

[0004] The present invention aims to provide a method for efficiently pushing business opportunities based on feature selection to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A method for efficiently pushing business opportunities based on feature selection includes the following steps:

[0007] Step 1: Integrate customer historical data and data collected by business personnel during customer contact (purchasing power, investment preferences, interests and hobbies, etc.), and pre-process the integrated data;

[0008] Step 2: Extract features related to insurance purchase intention from the integrated customer data set, and encode the extracted features to obtain a basic feature sequence, including customer purchase behavior characteristics (purchase frequency, purchase amount, etc.), basic attribute characteristics (age, gender, annual income, etc.), purchasing power characteristics, investment characteristics, and interest characteristics;

[0009] Step 3: Based on the feature data in the basic feature sequence, the 521 portrait features of the customer are integrated to build a customer portrait model, predict the customer's purchase intention, and divide the customer into different segments;

[0010] Step 4: Use the filtering feature selection method to screen the extracted features, find the feature subset that has the greatest impact on insurance purchase intention, sort the features according to the feature importance score, and assign weights to each feature to reflect its contribution in predicting customer purchase intention;

[0011] Step 5: Based on the customer portrait model and the weights assigned to each feature, calculate the willingness evaluation coefficient, screen and evaluate the purchase intention of the customer group, and provide personalized product recommendations and service suggestions to business personnel based on the specific portrait of the customer;

[0012] Step 6: Push the recommendation information to business personnel in the form of business opportunities, and track the actual effect of pushing business opportunities, including conversion rate and customer satisfaction indicators, to further evaluate the quality of business opportunity push.

[0013] A further improvement of the technical solution of the present invention is that in step 1, the customer historical data and the data collected by the business personnel during customer contact are specifically integrated as follows:

[0014] Export the customer's historical data from the database of the company's customer relationship management system, insurance sales system and claims system, including the customer's purchase record (purchased insurance product type, purchase time, purchase amount, policy status, etc.), age, gender, annual income and basic information of occupation;

[0015] Collect data collected by sales staff during customer contact from sales staff's work records and customer interview records, including customer purchasing power (income level, asset status, consumption habits, etc.), investment preferences (investment channels, investment period, risk preference, etc.) and hobbies (sports, travel, reading, etc.);

[0016] Pre-process the collected customer historical data and data collected by business personnel, including data cleaning and data standardization steps;

[0017] The pre-processed data is integrated, and the customer historical data and the data collected by business personnel are merged according to the customer number identification to form a complete customer data set.

[0018] A further improvement of the technical solution of the present invention is that in step 2, the process of obtaining the basic feature sequence is:

[0019] Perform feature analysis on the integrated customer data set to extract features related to insurance purchase intention, including customer purchase behavior features, basic attribute features, purchasing power features, investment features, and interest and hobby features;

[0020] For purchasing behavior characteristics, extract the customer's purchase frequency, purchase amount and purchase product type characteristics to analyze the customer's purchasing habits and preferences. For basic attribute characteristics, extract the customer's age, gender and annual income characteristics to analyze the customer's economic status. For purchasing power characteristics, extract the customer's consumption level and purchasing potential characteristics by analyzing the customer's purchase record and payment ability. For investment characteristics, extract the customer's investment behavior and investment preferences, including investment amount, investment field and investment style, and analyze the customer's investment awareness and risk tolerance. For hobby characteristics, extract the customer's hobby characteristics by analyzing the customer's social media data and interaction records.

[0021] The extracted features are encoded using classification coding, a unique coding value is assigned to each category, and the basic feature sequence is obtained by integration.

[0022] A further improvement of the technical solution of the present invention is that the 521 portrait features of the customer are divided into 5, 2 and 1, wherein 5 is 5 senses, indicating the customer's family structure, customer purchasing power (occupational housing consumption), insurance situation (what insurances are available, preferences), investment situation (investment preferences, financial situation) and interests and hobbies (what to like, interest preferences), 2 is 2 senses, specifically the customer's reason for purchase and the customer's reason for refusal to purchase, and 1 is 1-point communication, indicating that a purchase plan is provided to the customer.

[0023] A further improvement of the technical solution of the present invention is that in step 3, the process of predicting the customer's purchase intention is:

[0024] Integrate the feature data in the basic feature sequence and the "521 portrait feature" data, and perform data cleaning and standardization on the integrated data to facilitate subsequent analysis and modeling;

[0025] Integrate the "5 senses", "2 feelings" and "1 point of contact" features of the customer's "521 portrait features" into the basic feature sequence to form a comprehensive customer portrait feature set, ensure the accuracy and consistency of the features, and avoid introducing noise and redundant information;

[0026] Use the customer portrait feature set combined with the decision tree model to build a customer portrait model, process the feature data in the customer portrait feature set, use the customer portrait features as the input of the model, and the customer's purchase intention as the output of the model. Learn the relationship between customer portrait and purchase intention, and evaluate the performance of the model through cross-validation, accuracy, recall and other indicators;

[0027] Use the trained customer portrait model to predict the purchase intention of customer data, compare the prediction results with the actual purchase situation to verify the accuracy of the model, and conduct in-depth analysis of the prediction results to understand the distribution and trend of customer purchase intention and analyze the impact of different characteristics on purchase intention;

[0028] Based on the prediction results and customer portrait characteristics, we formulate standards and rules for customer segmentation and divide customers into different segments, namely, high purchase intention group, medium purchase intention group and low purchase intention group. We conduct in-depth analysis on the characteristics of each group to understand customers' purchasing behavior, needs and preferences.

[0029] A further improvement of the technical solution of the present invention is that in step 4, the process of assigning weights to each feature is as follows:

[0030] According to the type of features and the distribution of data, the variance selection method is used to analyze the relationship between features and target variables, and the importance score of each feature is calculated, where the target variable is the willingness to purchase insurance;

[0031] Preset the variance threshold of each feature, calculate the variance of each feature, and remove the features whose variance is less than the variance threshold. The values ​​of low-variance features are not very different, and their contribution to distinguishing samples is small.

[0032] The features are sorted according to the feature importance scores to obtain the feature importance ranking, where the sorted features are: annual income, purchase frequency, purchase amount, age, and gender;

[0033] According to the feature importance score and ranking results, the normalized variance of each feature is calculated, and a weight value is assigned to the annual income, purchase frequency, purchase amount, age and gender characteristics. The weight of annual income is set to 0.3, the weight of purchase frequency is set to 0.25, the weight of purchase amount is set to 0.2, the weight of age is set to 0.15, and the weight of gender is set to 0.1. The influence of different features on the willingness to purchase insurance is analyzed based on the feature weights.

[0034] A further improvement of the technical solution of the present invention is that in step 5, the calculation process of the willingness evaluation coefficient is:

[0035] The customer portrait model is used to analyze the customer's purchasing intention, and the feature values ​​in the customer portrait model are standardized according to the weights assigned to each feature to make them comparable under the same dimension.

[0036] Multiply the standardized feature value of each customer by the corresponding feature weight to obtain the weighted feature value, and sum the weighted feature values ​​of the customers to obtain the willingness evaluation coefficient, and then screen and evaluate the purchase intention of customers in the customer group;

[0037] According to the size of the willingness evaluation coefficient, the customer groups are grouped according to the standards of high purchase intention group, medium purchase intention group and low purchase intention group, and the screened customer groups are evaluated to understand the distribution of their purchase intentions;

[0038] Based on the customer's specific portrait, analyze their purchasing behavior, needs and preferences, formulate personalized product recommendation strategies, and then provide business personnel with personalized product recommendation plans, including recommended product types, insurance amounts and coverage period information, to assist business personnel in connecting with corresponding customer groups.

[0039] A further improvement of the technical solution of the present invention is that the calculation formula of the willingness evaluation coefficient is:

[0040]

[0041] Among them, W is the willingness evaluation coefficient, ω i is the weight of the i-th feature, f i is the standardized value of the ith feature, n is the total number of features, α is an adjustment parameter used to control the rate of exponential decay, and the value range of α is between 0 and 1. It should be noted that the value range of W is between 0 and 1. When the weighted values ​​of all features are close to 0, W is close to 0, indicating that the customer's willingness to buy is low. When the weighted values ​​of all features are close to 1, W is close to 1, indicating that the customer's willingness to buy is high.

[0042] Set multiple evaluation thresholds for each of the multiple purchase intention groups, specifically:

[0043] The evaluation threshold range of the high purchase intention group is: W H ≤W<1; customers in this group have a high interest in and willingness to purchase insurance products and are the focus of insurance companies’ marketing and maintenance;

[0044] The evaluation threshold range of the medium purchase intention group is: W M ≤W <W H ; Customers in this group have a certain interest in insurance products, but their willingness to purchase has not yet reached a high level. Insurance companies can tap into their potential purchasing needs through precision marketing and personalized services;

[0045] The evaluation threshold range of the low purchase intention group is: 0 <W<W M ; Customers in this group have a low demand for insurance products and are not very willing to buy. Insurance companies need to analyze the reasons and adopt corresponding strategies to increase their willingness to buy;

[0046] Among them, W is the willingness evaluation coefficient, W His the lower threshold of the high purchase intention group and the upper threshold of the medium purchase intention group, W M is the lower threshold of the medium purchase intention group and the upper threshold of the low purchase intention group, W H =0.8,W M =0.4.

[0047] A further improvement of the technical solution of the present invention is that in step 6, the evaluation process of the quality of business opportunity push is:

[0048] Organize the personalized product recommendation plan based on the customer profile and willingness evaluation coefficient into business opportunity information, which includes the customer's basic information (name, contact information, etc.), the recommended insurance product type, insurance amount, protection period information, and the reason and basis for the recommendation, and push the confirmed business opportunity information to the business personnel through the internal customer relationship management system;

[0049] Sales personnel follow up based on business opportunity information and record customer feedback and conversion status, including whether the deal is completed, transaction time, and transaction amount. After the deal is completed, customer satisfaction data is collected through questionnaires. Customer satisfaction data covers product quality, service experience, and cost-effectiveness.

[0050] Calculate the conversion rate and customer satisfaction score of the pushed business opportunities, and then calculate the business opportunity quality evaluation coefficient based on the conversion rate and customer satisfaction indicators to conduct a comprehensive assessment of the quality of business opportunity push;

[0051] Analyze the results of the business opportunity quality evaluation coefficient, understand the advantages and disadvantages of business opportunity push, and based on the evaluation results, put forward suggestions for improving the quality of business opportunity push, feedback the evaluation results and improvement suggestions to all aspects of business opportunity push, continuously optimize the process and strategy of business opportunity push, and continuously improve the quality and effect of business opportunity push.

[0052] A further improvement of the technical solution of the present invention is that the calculation formula of the business opportunity quality evaluation coefficient is:

[0053]

[0054] Among them, Q is the business opportunity quality evaluation coefficient, C is the business opportunity conversion rate, which indicates the proportion of business opportunities converted into actual sales. The higher the conversion rate, the better the sales effect of the business opportunity and the greater the contribution to the quality of the business opportunity. S is the customer satisfaction score, which indicates the customer's satisfaction with the purchased insurance products and services. The higher the customer satisfaction score, the better the customer experience brought by the business opportunity and the greater the contribution to the quality of the business opportunity. The conversion rate is adjusted so that the business opportunity quality evaluation coefficient tends to be stable when the conversion rate is high, to avoid the business opportunity quality evaluation coefficient being too large due to excessive conversion rate. It should be noted that the value range of Q is between 0 and 1. When the business opportunity conversion rate C and the customer satisfaction score S are both 0, Q is close to 0, indicating that the business opportunity quality is poor; when the business opportunity conversion rate C and the customer satisfaction score S are close to the maximum value, Q is close to 1, indicating that the business opportunity quality is good.

[0055] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0056] 1. The present invention provides a method for efficiently pushing business opportunities based on feature selection. By collecting and analyzing multi-dimensional data of customers, an accurate customer portrait is constructed. Based on the accurate matching of the customer portrait, the accuracy of the push is greatly improved, invalid push is reduced, and customer satisfaction with the push is improved. Moreover, since the content of the push is highly matched with customer needs, customers are more likely to have a willingness to buy, thereby improving the conversion rate. In addition, accurate push can reduce the investment in invalid customers, invest more resources in potential high-value customers, and achieve the optimal allocation of resources.

[0057] 2. The present invention provides a method for efficiently pushing business opportunities based on feature selection. By analyzing the willingness evaluation coefficient, it ensures that the pushed content is synchronized with customer needs and improves the timeliness of the push. Since the pushed content is closer to customer needs, the customer experience is improved, which helps to enhance the customer's favorability towards the company. By analyzing the changes in the willingness evaluation coefficient, problems in the push strategy can be discovered in time, and optimization and adjustment can be made to improve the push effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0059] Figure 1 is a flow chart of the method of the present invention;

[0060] Figure 2 is a calculation flow chart of the willingness evaluation coefficient of the present invention;

[0061] Figure 3 This is a flow chart for evaluating the quality of business opportunity push according to the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Embodiment 1, as Figure 1 As shown, the present invention provides a method for efficiently pushing business opportunities based on feature selection, comprising the following steps:

[0064] Step 1: Integrate customer historical data and data collected by sales personnel during customer contact (purchasing power, investment preferences, interests and hobbies, etc.), and pre-process the integrated data. Export customer historical data from the database of the company's customer relationship management system, insurance sales system and claims system, including customer purchase records (purchased insurance product type, purchase time, purchase amount, policy status, etc.), age, gender, annual income and basic information of occupation. Collect data collected by sales personnel during customer contact from sales personnel's work records and customer interview records, including customer purchasing power (income level, asset status, consumption habits, etc.), investment preferences (investment channels, investment period, risk preference, etc.) and Hobbies (sports, travel, reading, etc.), pre-processing the collected customer historical data and data collected by business personnel, including data cleaning and data standardization processing steps, in which, through data cleaning, duplicate records in the data set are checked and deleted to ensure the uniqueness of the data, missing data are filled or deleted according to the characteristics of the data and business needs, errors in the data are checked and corrected, and data from different sources and formats are unified through data standardization, and numerical data are converted into a unified unit of measurement. The pre-processed data is integrated, and the customer historical data and the data collected by business personnel are merged according to the customer number identifier to form a complete customer data set;

[0065] Step 2: Extract features related to the willingness to purchase insurance from the integrated customer data set, and encode the extracted features to obtain a basic feature sequence, including the customer's purchase behavior characteristics (purchase frequency, purchase amount, etc.), basic attribute characteristics (age, gender, annual income, etc.), purchasing power characteristics, investment characteristics, and interest and hobby characteristics. Perform feature analysis on the integrated customer data set to extract features related to the willingness to purchase insurance, including the customer's purchase behavior characteristics, basic attribute characteristics, purchasing power characteristics, investment characteristics, and interest and hobby characteristics. For the purchase behavior characteristics, extract the customer's purchase frequency, purchase amount, and purchase product type characteristics, and analyze the customer's purchase habits. For basic attribute characteristics, extract the customer's age, gender and annual income characteristics, analyze the customer's economic status, for purchasing power characteristics, extract the customer's consumption level and purchasing potential characteristics by analyzing the customer's purchase record and payment ability, for investment characteristics, extract the customer's investment behavior and investment preferences, including investment amount, investment field and investment style, analyze the customer's investment awareness and risk tolerance, for hobbies characteristics, extract the customer's hobbies characteristics by analyzing the customer's social media data and interaction records, and encode the extracted features using classification coding, assign a unique coding value to each category, and integrate to obtain the basic feature sequence;

[0066] Furthermore, for the coding of purchase behavior characteristics, the purchase frequency is segmented and coded according to the number of purchases, which is divided into "low frequency", "medium frequency" and "high frequency", coded as 0, 1, and 2 respectively. "Low frequency" is set as the number of purchases within a year is less than or equal to 2 times, "medium frequency" is 3-5 times, and "high frequency" is greater than 5 times. The purchase amount is segmented and coded according to the amount range, which is divided into "low consumption", "medium consumption" and "high consumption", coded as 0, 1, and 2 respectively. "Low consumption" is set as the purchase amount is less than or equal to 5,000 yuan, "medium consumption" is 5,000-20,000 yuan, and "high consumption" is greater than 20,000 yuan. The type of purchased product is coded according to the product type, which is divided into "life insurance", "accident insurance" and "health insurance", coded as 0, 1, and 2 respectively;

[0067] For the basic attribute feature coding, age is coded by age group, which are "18-24 years old", "25-34 years old", "35-44 years old", "45-54 years old" and "over 55 years old", coded as 0, 1, 2, 3, 4 respectively. For gender, male = 1, female = 0. Annual income is coded by income range, which are divided into "low income", "middle income" and "high income", coded as 0, 1, 2 respectively. "Low income" is set as annual income less than or equal to 50,000 yuan, "middle income" is 50,000-200,000 yuan, and "high income" is greater than 200,000 yuan.

[0068] For the coding of purchasing power characteristics, consumption level was coded according to consumption level, divided into “low level”, “middle level” and “high level”, coded as 0, 1, and 2 respectively, and purchasing potential was coded according to the size of potential, divided into “low potential”, “middle potential” and “high potential”, coded as 0, 1, and 2 respectively;

[0069] For the coding of investment characteristics, the investment amount is segmented and coded according to the amount range, and is divided into "low investment", "medium investment" and "high investment", coded as 0, 1, and 2 respectively. "Low investment" is set as an investment amount less than or equal to 100,000 yuan, "medium investment" is 100,000-500,000 yuan, and "high investment" is greater than 500,000 yuan. The investment field is coded by field, and is divided into "stocks", "funds" and "real estate", coded as 0, 1, and 2 respectively. The investment style is coded by style, and is divided into "conservative", "sound" and "aggressive", coded as 0, 1, and 2 respectively.

[0070] For the feature coding of interests and hobbies, interests and hobbies are coded according to interest points, including "sports enthusiasts", "travel enthusiasts" and "reading enthusiasts", which are coded as 0, 1, and 2 respectively. For multiple interest points, multi-value coding is used for processing;

[0071] Step 3. Based on the feature data in the basic feature sequence, the customer's 521 portrait features are integrated to build a customer portrait model, predict the customer's purchase intention, and divide the customer into different segments. The customer's 521 portrait features are divided into 5, 2 and 1, among which 5 is 5 senses, representing the customer's family structure, customer purchasing power (occupational housing consumption), insurance situation (what insurances are available, preferences), investment situation (investment preferences, financial situation) and hobbies (what do you like, interest preferences), 2 is 2 senses, specifically the customer's purchase reasons and customer refusal reasons, 1 is 1 point, indicating that a purchase plan is provided to the customer, the feature data in the basic feature sequence and the "521 portrait feature" data are integrated, and the integrated data is cleaned and standardized for subsequent analysis and modeling, and the "5 senses", "2 senses" and "1 point" features in the customer's "521 portrait features" are integrated into the basic feature sequence to form a comprehensive customer portrait feature set to ensure the accuracy and consistency of the features. , avoid introducing noise and redundant information, use the customer portrait feature set combined with the decision tree model to build a customer portrait model, process the feature data in the customer portrait feature set, use the customer portrait features as the input of the model, and the customer's purchase intention as the output of the model, and learn the relationship between customer portrait and purchase intention, evaluate the performance of the model through cross-validation, accuracy, recall and other indicators, use the trained customer portrait model to predict the purchase intention of customer data, compare the prediction results with the actual purchase situation to verify the accuracy of the model, and conduct in-depth analysis of the prediction results to understand the distribution and trend of customer purchase intention, analyze the impact of different features on purchase intention, and formulate customer segmentation standards and rules based on the prediction results and customer portrait features, and divide customers into different segmented groups, namely, high purchase intention group, medium purchase intention group and low purchase intention group, conduct in-depth analysis of the characteristics of each group, and understand customers' purchasing behavior, needs and preferences;

[0072] Step 4: Use the filtering feature selection method to screen the extracted features, find the feature subset that has the greatest impact on the willingness to buy insurance, and sort the features according to the feature importance score, assign weights to each feature to reflect its contribution in predicting the customer's willingness to buy, and use the variance selection method to analyze the relationship between the feature and the target variable according to the type of feature and the distribution of the data, and calculate the importance score of each feature. The target variable is the willingness to buy insurance, and the variance threshold of each feature is preset, and the variance of each feature is calculated. Features with variances less than the variance threshold are removed. The values ​​of low-variance features are not very different, and their contribution to distinguishing samples is relatively small. Small, sort the features according to the feature importance scores to get the feature importance ranking, where the sorted features are: annual income, purchase frequency, purchase amount, age and gender. According to the feature importance scores and sorting results, calculate the normalized variance of each feature, and assign a weight value to the annual income, purchase frequency, purchase amount, age and gender features. Set the weight of annual income to 0.3, the weight of purchase frequency to 0.25, the weight of purchase amount to 0.2, the weight of age to 0.15, and the weight of gender to 0.1. Analyze the influence of different features on the willingness to buy insurance according to the feature weights;

[0073] Step 5: Based on the customer portrait model and the weights assigned to each feature, calculate the willingness evaluation coefficient, screen and evaluate the purchase intention of the customer group, and provide personalized product recommendations and service suggestions to business personnel based on the specific portrait of the customer;

[0074] Step 6: Push the recommendation information to business personnel in the form of business opportunities, and track the actual effect of pushing business opportunities, including conversion rate and customer satisfaction indicators, to further evaluate the quality of business opportunity push.

[0075] Embodiment 2, as Figure 2 , Figure 3 As shown, based on Example 1, the present invention provides a technical solution: Preferably, in step 5, the calculation process of the willingness evaluation coefficient is:

[0076] Use the customer portrait model to analyze the customer's willingness to buy, and combine the weights assigned to each feature to standardize the eigenvalues ​​in the customer portrait model to make them comparable under the same dimension. Multiply the standardized eigenvalue of each customer by the corresponding eigenweight to obtain the weighted eigenvalue, and sum the weighted eigenvalues ​​of the customer to obtain the willingness evaluation coefficient. Then, screen and evaluate the purchase intention of customers in the customer group. According to the size of the willingness evaluation coefficient, group the customer group according to the standards of high purchase intention group, medium purchase intention group and low purchase intention group, and evaluate the screened customer groups to understand the distribution of their purchase intention. According to the specific portrait of the customer, analyze their purchase behavior, needs and preferences, formulate personalized product recommendation strategies, and then provide business personnel with personalized product recommendation plans, including recommended product types, insurance amounts and protection period information, to assist business personnel in connecting with corresponding customer groups.

[0077] Furthermore, the calculation formula of the willingness evaluation coefficient is:

[0078]

[0079] Among them, W is the willingness evaluation coefficient, ω i is the weight of the i-th feature, f i is the normalized value of the ith feature, n is the total number of features, α is an adjustment parameter used to control the rate of exponential decay, and the value range of α is between 0 and 1. It represents the weighted sum of all features, reflecting the overall performance of the customer on all features. is the Euclidean norm of the weighted eigenvalue, which is used to normalize the result of the summation function to ensure that the value range of the willingness evaluation coefficient is between 0 and 1. Used to adjust the value trend of the willingness evaluation coefficient, when the average value of the weighted eigenvalue When it is close to 1, the value of the logarithmic function is close to 0, and the value of the exponential function is close to 1; when the average value of the weighted eigenvalue is close to 0, the value of the logarithmic function is close to negative infinity, and the value of the exponential function is close to 0. The willingness evaluation coefficient will increase with the increase of the average value of the weighted eigenvalue, but the rate of increase will be controlled by exponential decay. It should be noted that the value range of W is between 0 and 1. When the weighted values ​​of all features are close to 0, W is close to 0, indicating that the customer's willingness to buy is low. When the weighted values ​​of all features are close to 1, W is close to 1, indicating that the customer's willingness to buy is high. As the average value of the weighted eigenvalue increases, W will increase, but the rate of increase will be controlled by exponential decay, that is, the increase of W will gradually slow down;

[0080] Furthermore, multiple evaluation thresholds are set corresponding to multiple purchase intention groups, specifically:

[0081] The evaluation threshold range of the high purchase intention group is: W H ≤W<1; customers in this group have a high interest in and willingness to purchase insurance products and are the focus of insurance companies’ marketing and maintenance;

[0082] The evaluation threshold range of the medium purchase intention group is: W M ≤W <W H ; Customers in this group have a certain interest in insurance products, but their willingness to purchase has not yet reached a high level. Insurance companies can tap into their potential purchasing needs through precision marketing and personalized services;

[0083] The evaluation threshold range for the low purchase intention group is: 0 <W<W M ; Customers in this group have a low demand for insurance products and are not very willing to buy. Insurance companies need to analyze the reasons and adopt corresponding strategies to increase their willingness to buy;

[0084] Among them, W is the willingness evaluation coefficient, W H is the lower threshold of the high purchase intention group and the upper threshold of the medium purchase intention group, W M is the lower threshold of the medium purchase intention group and the upper threshold of the low purchase intention group, W H =0.8,W M =0.4;

[0085] In step 6, the evaluation process of the quality of business opportunity push is as follows:

[0086] The personalized product recommendation plan formulated according to the customer portrait and willingness evaluation coefficient is organized into business opportunity information. The business opportunity information includes the customer's basic information (name, contact information, etc.), the recommended insurance product type, the insured amount, the protection period information, and the reasons and basis for the recommendation. The confirmed business opportunity information is pushed to the business personnel through the internal customer relationship management system. The business personnel follow up according to the business opportunity information and record the customer's feedback and conversion status. The conversion status includes whether the transaction is completed, the transaction time and the transaction amount information. After the business opportunity is completed, the customer satisfaction data is collected through questionnaires. The customer satisfaction data covers product quality, service experience and cost-effectiveness. The conversion rate and customer satisfaction score of the pushed business opportunities are calculated, and then the business opportunity quality evaluation coefficient is calculated in combination with the conversion rate and customer satisfaction indicators. The quality of business opportunity push is comprehensively evaluated, the results of the business opportunity quality evaluation coefficient are analyzed, the advantages and disadvantages of business opportunity push are understood, and suggestions for improving the quality of business opportunity push are put forward based on the evaluation results. The evaluation results and improvement suggestions are fed back to each link of business opportunity push, and the process and strategy of business opportunity push are continuously optimized to continuously improve the quality and effect of business opportunity push.

[0087] Furthermore, the calculation formula of the business opportunity quality evaluation coefficient is:

[0088]

[0089] Among them, Q is the business opportunity quality evaluation coefficient, C is the business opportunity conversion rate, which indicates the proportion of business opportunities converted into actual sales. The higher the conversion rate, the better the sales effect of the business opportunity and the greater the contribution to the quality of the business opportunity. S is the customer satisfaction score, which indicates the customer's satisfaction with the purchased insurance products and services. The higher the customer satisfaction score, the better the customer experience brought by the business opportunity and the greater the contribution to the quality of the business opportunity. The conversion rate is adjusted so that the business opportunity quality evaluation coefficient tends to be stable when the conversion rate is high, avoiding an excessively large business opportunity quality evaluation coefficient due to an excessively high conversion rate. It should be noted that the value range of Q is between 0 and 1. When the business opportunity conversion rate C and the customer satisfaction score S are both 0, Q is close to 0, indicating that the business opportunity quality is poor; when the business opportunity conversion rate C and the customer satisfaction score S are close to the maximum value, Q is close to 1, indicating that the business opportunity quality is good. With the increase of the business opportunity conversion rate C and the customer satisfaction score S, the business opportunity quality evaluation coefficient Q will increase, but the rate of increase will gradually slow down and tend to a stable value. This is in line with the actual situation of business opportunity quality evaluation, that is, the improvement of business opportunity quality requires comprehensive consideration of the balanced development of conversion rate and customer satisfaction.

[0090] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for efficiently pushing business opportunities based on feature selection, characterized in that: The following steps are involved: Step 1: Integrate customer historical data and data collected by business personnel during customer contact, and pre-process the integrated data; Step 2: Extract features related to insurance purchase intention from the integrated customer data set, and encode the extracted features to obtain a basic feature sequence; Step 3: Based on the feature data in the basic feature sequence, the 521 portrait features of the customer are integrated to build a customer portrait model, predict the customer's purchase intention, and divide the customer into different segments; Step 4: Use the filtering feature selection method to screen the extracted features, sort the features according to the feature importance score, and assign weights to each feature; Step 5: Based on the customer portrait model and the weights assigned to each feature, calculate the willingness evaluation coefficient, screen and evaluate the purchase intention of the customer group, and provide personalized product recommendations and service suggestions to business personnel based on the specific portrait of the customer; Step 6: Push the recommendation information to business personnel in the form of business opportunities, and track the actual effect of pushing business opportunities, including conversion rate and customer satisfaction indicators, to further evaluate the quality of business opportunity push.

2. The method for efficiently pushing business opportunities based on feature selection according to claim 1, characterized in that: In step 1, the customer historical data and the data collected by the business personnel during customer contact are integrated in the following specific process: Export the customer's historical data from the databases of the company's customer relationship management system, insurance sales system and claims system, including basic information about the customer's purchase record, age, gender, annual income and occupation; Collect data collected by sales staff during customer contact from sales staff's work records and customer interview records, including customer purchasing power, investment preferences, and hobbies; Pre-process the collected customer historical data and data collected by business personnel, including data cleaning and data standardization steps; The pre-processed data is integrated, and the customer historical data and the data collected by business personnel are merged according to the customer number identification to form a complete customer data set.

3. The method for efficiently pushing business opportunities based on feature selection according to claim 2 is characterized in that: In step 2, the process of obtaining the basic feature sequence is as follows: Perform feature analysis on the integrated customer data set to extract features related to insurance purchase intention, including customer purchase behavior features, basic attribute features, purchasing power features, investment features, and interest and hobby features; For purchasing behavior characteristics, extract the customer's purchase frequency, purchase amount and purchase product type characteristics to analyze the customer's purchasing habits and preferences. For basic attribute characteristics, extract the customer's age, gender and annual income characteristics to analyze the customer's economic status. For purchasing power characteristics, extract the customer's consumption level and purchasing potential characteristics by analyzing the customer's purchase record and payment ability. For investment characteristics, extract the customer's investment behavior and investment preferences, including investment amount, investment field and investment style, and analyze the customer's investment awareness and risk tolerance. For hobby characteristics, extract the customer's hobby characteristics by analyzing the customer's social media data and interaction records. The extracted features are encoded using classification coding, a unique coding value is assigned to each category, and the basic feature sequence is obtained by integration.

4. The method for efficiently pushing business opportunities based on feature selection according to claim 3 is characterized in that: The 521 portrait features of the customer are divided into 5, 2 and 1, among which 5 stands for 5 senses, representing the customer's family structure, purchasing power, insurance situation, investment situation and interests and hobbies; 2 stands for 2 senses, specifically the customer's reasons for purchase and the reasons for customer refusal to purchase; 1 stands for 1 point connection, indicating providing the customer with a purchase plan.

5. The method for efficiently pushing business opportunities based on feature selection according to claim 4 is characterized in that: In step 3, the process of predicting customer purchase intention is as follows: Integrate the feature data in the basic feature sequence and the "521 portrait feature" data, and perform data cleaning and standardization on the integrated data; Integrate the "5 senses", "2 feelings" and "1 point of contact" features of the customer's "521 portrait features" into the basic feature sequence to form a comprehensive customer portrait feature set; Use the customer portrait feature set combined with the decision tree model to build a customer portrait model, process the feature data in the customer portrait feature set, use the customer portrait features as the input of the model, and the customer's purchase intention as the output of the model, and learn the relationship between customer portrait and purchase intention; Use the trained customer portrait model to predict the purchase intention of customer data, compare the prediction results with the actual purchase situation to verify the accuracy of the model, and conduct in-depth analysis of the prediction results to understand the distribution and trend of customer purchase intention and analyze the impact of different characteristics on purchase intention; Based on the prediction results and customer portrait characteristics, standards and rules for customer segmentation are formulated to divide customers into different segments, namely, high purchase intention group, medium purchase intention group and low purchase intention group.

6. The method for efficiently pushing business opportunities based on feature selection according to claim 5, characterized in that: In step 4, the process of assigning weights to each feature is as follows: According to the type of features and the distribution of data, the variance selection method is used to analyze the relationship between features and target variables, and the importance score of each feature is calculated, where the target variable is the willingness to purchase insurance; Preset the variance threshold of each feature, calculate the variance of each feature, and remove features whose variance is less than the variance threshold; The features are sorted according to the feature importance scores to obtain the feature importance ranking, where the sorted features are: annual income, purchase frequency, purchase amount, age, and gender; According to the feature importance score and ranking results, the normalized variance of each feature is calculated, and a weight value is assigned to the annual income, purchase frequency, purchase amount, age and gender characteristics. The weight of annual income is set to 0.3, the weight of purchase frequency is set to 0.25, the weight of purchase amount is set to 0.2, the weight of age is set to 0.15, and the weight of gender is set to 0.

1. The influence of different features on the willingness to purchase insurance is analyzed based on the feature weights.

7. The method for efficiently pushing business opportunities based on feature selection according to claim 6, characterized in that: In step 5, the calculation process of the willingness evaluation coefficient is: Use the customer portrait model to analyze the customer's purchasing intention, and standardize the feature values ​​in the customer portrait model based on the weights assigned to each feature; Multiply the standardized feature value of each customer by the corresponding feature weight to obtain the weighted feature value, and sum the weighted feature values ​​of the customers to obtain the willingness evaluation coefficient, and then screen and evaluate the purchase intention of customers in the customer group; According to the size of the willingness evaluation coefficient, the customer groups are grouped according to the standards of high purchase intention group, medium purchase intention group and low purchase intention group, and the screened customer groups are evaluated to understand the distribution of their purchase intentions; Based on the customer's specific portrait, analyze their purchasing behavior, needs and preferences, formulate personalized product recommendation strategies, and then provide business personnel with personalized product recommendation plans, including recommended product types, insurance amounts and coverage period information, to assist business personnel in connecting with corresponding customer groups.

8. The method for efficiently pushing business opportunities based on feature selection according to claim 7 is characterized in that: The calculation formula of the willingness evaluation coefficient is: Among them, W is the willingness evaluation coefficient, ω i is the weight of the i-th feature, f i is the standardized value of the ith feature, n is the total number of features, α is an adjustment parameter used to control the rate of exponential decay. It should be noted that the value range of W is between 0 and 1. When the weighted values ​​of all features are close to 0, W is close to 0, indicating that the customer's willingness to buy is low. When the weighted values ​​of all features are close to 1, W is close to 1, indicating that the customer's willingness to buy is high. Set multiple evaluation thresholds for each of the multiple purchase intention groups, specifically: The evaluation threshold range of the high purchase intention group is: W H ≤W<1; The evaluation threshold range of the medium purchase intention group is: W M ≤W <W H ; The evaluation threshold range of the low purchase intention group is: 0 <W<W M ; Among them, W is the willingness evaluation coefficient, W H is the lower threshold of the high purchase intention group and the upper threshold of the medium purchase intention group, W M is the lower threshold of the medium purchase intention group and the upper threshold of the low purchase intention group, W H =0.8,W M =0.

4.

9. The method for efficiently pushing business opportunities based on feature selection according to claim 8, characterized in that: In step 6, the evaluation process of the quality of business opportunity push is as follows: Organize the personalized product recommendation plan based on the customer profile and willingness evaluation coefficient into business opportunity information, which includes the customer's basic information, recommended insurance product type, insurance amount, protection period information, and reasons and basis for the recommendation, and push the confirmed business opportunity information to the business personnel through the internal customer relationship management system; Sales personnel follow up based on business opportunity information and record customer feedback and conversion status, including whether the deal is completed, transaction time and transaction amount. After the business opportunity is completed, customer satisfaction data is collected through questionnaires; Calculate the conversion rate and customer satisfaction score of the pushed business opportunities, and then calculate the business opportunity quality evaluation coefficient based on the conversion rate and customer satisfaction indicators to conduct a comprehensive assessment of the quality of business opportunity push; Analyze the results of the business opportunity quality evaluation coefficient, understand the advantages and disadvantages of business opportunity push, and based on the evaluation results, put forward suggestions for improving the quality of business opportunity push, feed back the evaluation results and improvement suggestions to all aspects of business opportunity push, and continuously optimize the process and strategy of business opportunity push.

10. The method for efficiently pushing business opportunities based on feature selection according to claim 9, characterized in that: The calculation formula of the business opportunity quality evaluation coefficient is: Among them, Q is the business opportunity quality evaluation coefficient, C is the business opportunity conversion rate, and S is the customer satisfaction score. It should be noted that the value range of Q is between 0 and 1. When the business opportunity conversion rate C and the customer satisfaction score S are both 0, Q is close to 0, indicating that the business opportunity quality is poor; when the business opportunity conversion rate C and the customer satisfaction score S are both close to the maximum value, Q is close to 1, indicating that the business opportunity quality is good.

Images