Potential customer mining and active registration guidance method and system based on big data analysis

Through the combination of big data analysis, logistic regression and OCSVM model, potential customers are identified and guided, which solves the problem of insufficient recognition capabilities of traditional models under data imbalance, and improves the accuracy of customer mining and marketing efficiency.

CN118569906BActive Publication Date: 2024-12-06JIANGSU HUANXUN INFORMATION TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410998667.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-12-06
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

When processing customer data, traditional binary classification models are difficult to effectively identify potential customers due to data imbalance, resulting in a decrease in prediction accuracy and generalization ability.

Method used

Through big data analysis, customer data is collected and cleaned, customer groups are divided through similarity analysis, preset impact index, combined with logistic regression and OCSVM model, identify potential customers and update impact index, and design personalized guidance strategies.

Benefits of technology

It improves the recognition rate of potential customers and the targetedness and efficiency of marketing activities, significantly improves the accuracy and generalization capabilities of data processing, and solves the problem of data imbalance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118569906B_ABST
    Figure CN118569906B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and system for potential customer mining and active registration guidance based on big data analysis, including the following steps: collecting and cleaning customer data, dividing them into two categories: products that have been purchased and products that have not been purchased; dividing customers into different customer groups through similarity analysis, each customer group has similar purchasing behaviors and preferences; after presetting the influence index for each customer group, using the logistic regression model to identify the potential customer group that has the intention to purchase but has not purchased; using the OCSVM model to calculate the abnormal scores of these customers, adjusting the customer group influence index according to the score, and re-entering the model to output the second potential customer group; sending customized financial product or service information to these customers through the guidance channel, guiding them to register and trade, improving the recognition rate of potential customers, and ensuring the pertinence and efficiency of marketing activities. It effectively solves the problem of category imbalance in big data analysis and improves the accuracy and generalization ability of data processing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, specifically to a data processing method and system suitable for commercial, financial, management, supervision or prediction purposes, and in particular to a potential customer mining and active registration guidance method and system based on big data analysis. Background Art

[0002] With the continuous advancement of big data and artificial intelligence technologies, financial institutions have ushered in new development opportunities in customer insights and marketing strategies. Traditional customer analysis methods mainly rely on limited data sources and static analysis models, which are difficult to adapt to the rapid changes in the market and the diversified needs of customers.

[0003] In actual big data analysis applications, the customer data involved includes both customers who have purchased financial products and customers who have not purchased financial products. However, a common problem is that the number of customers who have not purchased financial products is much larger than the number of customers who have purchased them. This imbalance in data distribution leads to a significant difference in the amount of data between customers who have purchased and those who have not purchased.

[0004] In the financial field, binary classification models are a common data analysis technique used to identify potential customers from a large amount of customer data. However, when the data used as model input is biased, traditional binary classification models cannot achieve ideal results. In particular, when the number of samples in one category in the data set is much larger than that of the other category, the model is overly biased towards the majority category during the learning process, resulting in a decrease in the ability to identify the minority category. In other words, the model will over-learn about customers who have not purchased financial products, thereby affecting its prediction accuracy and generalization ability for potential customers.

[0005] Therefore, it is necessary to provide a method and system for potential customer mining and active registration guidance based on big data analysis to solve the above technical problems. Summary of the invention

[0006] The present invention overcomes the deficiencies of the prior art and provides a method and system for potential customer mining and active registration guidance based on big data analysis.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a method for potential customer mining and active registration guidance based on big data analysis, comprising the following steps:

[0008] S1. Collect a large amount of customer data, including customer data of products that have been purchased and those that have not been purchased, and clean and standardize the collected customer data to obtain a first data set;

[0009] S2. Performing similarity analysis on the customers in the first data set, identifying customer similarities between a number of customers in the first data set, and dividing the customers in the first data set into different customer groups based on the similarity analysis results, wherein a number of customers in the same customer group have similar purchasing behaviors and preferences;

[0010] S3, presetting an influence index for each customer group to obtain a second data set;

[0011] S4. Input the second data set into the logistic regression model, and output a first potential customer group, where the first potential customers are customers who have purchase intentions but have not yet made a purchase. The logistic regression model marks these customers as abnormal.

[0012] S5, inputting the first potential customer group into the OCSVM model, calculating the anomaly scores of all the first potential customers, modifying the preset influence index of the customer group to which these first potential customers belong according to the anomaly scores of the first potential customers, re-inputting the updated second data set into the logistic regression model, and outputting the second potential customer group;

[0013] S6. Send guidance information to the second potential customer group through the guidance channel, and guide the second potential customer group to register and trade through customized products or services.

[0014] In a preferred embodiment of the present invention, the similarity analysis in S2 comprises the following steps:

[0015] S21, selecting features related to customer purchasing behavior and preferences from the first data set;

[0016] S22, performing similarity calculation on the features selected in step S21, and calculating the similarity score between each pair of customers;

[0017] S23. Construct the calculated similarity scores into a similarity matrix, where each element in the matrix represents the similarity between a pair of customers.

[0018] In a preferred embodiment of the present invention, in S22, the selected features are first standardized before similarity calculation, using a Z-score standardization method: ,in is the original data, is the mean of the feature, is the standard deviation;

[0019] The standardized data is obtained by calculation, and then the Euclidean distance between each pair of customers is calculated: ,in and are the normalized feature vectors of the two customers, yes The characteristic value of yes The characteristic value of .

[0020] In a preferred embodiment of the present invention, the non-numeric data in the feature is converted into numerical data through binary encoding, one-hot encoding or label encoding method.

[0021] In a preferred embodiment of the present invention, the calculation of the anomaly score for the first potential customer in S5 includes the following steps:

[0022] S51. Use the python framework to build the OCSVM model;

[0023] S52, dividing the purchased customer data in the second data set into a training set and a test set to train the OCSVM model. After the training, a decision boundary is obtained in the OCSVM model, and the decision boundary includes the purchased customer data in the second data set and excludes the abnormal points;

[0024] S53. Input the first potential customer into the trained OCSVM model, and calculate the anomaly score of each potential customer. The anomaly score is the distance from the data point of the potential customer mapped to the decision hyperplane to the decision boundary. The higher the anomaly score, the greater the difference in characteristics between the first potential customer and the group of purchased customers.

[0025] In a preferred embodiment of the present invention, the updating of the preset impact index is specifically as follows:

[0026] Divide the first potential customer group into original customer groups and calculate the average anomaly score for each customer group ;

[0027] Calculate the average anomaly score for all customer groups , is the total number of customers;

[0028] Calculate the customer impact index adjustment factor , is the preset influence index of customer group j;

[0029] Add the impact index adjustment factor to the preset impact index to obtain the unnormalized new impact index for each customer group, and then calculate the reciprocal of the sum of all unnormalized new impact indices. ;

[0030] Calculate the normalized new impact index for each customer group ;

[0031] The normalized new impact index of each customer group Update to each customer group in the second dataset.

[0032] In a preferred embodiment of the present invention, in S6, the second potential customers are segmented into different groups based on the purchase behavior, preference and geographic location information of the second potential customers output by S5, and targeted guidance strategies are formulated, including providing personalized content, discounts and services.

[0033] A potential customer mining and active registration guidance system based on big data analysis, using any of the above-mentioned potential customer mining and active registration guidance methods based on big data analysis, including: a data collection module, a data preprocessing module, a similarity analysis module, a grouping module, an influence index preset module, a primary prediction module, an influence index update module, a secondary prediction module and an active registration guidance module;

[0034] The data collection module collects a large amount of customer data from multiple channels;

[0035] The data preprocessing module is used to perform data cleaning and standardization processing on the original customer data collected by the data collection module, and output a first data set;

[0036] The similarity analysis module is used to perform similarity analysis on the customer information in the first data set and calculate the similarity between each pair of customers;

[0037] The grouping module divides the customers in the first data set into different customer groups according to the similarity analysis result, and several customers in the same customer group have similar purchasing behaviors and preferences;

[0038] The influence index preset module presets an influence index for each customer group and combines them into a second data set as input to the prediction module;

[0039] The primary prediction module establishes a logistic regression model, inputs the second data set, analyzes and predicts whether the customers in the second data set are potential customers, and outputs a first potential customer group;

[0040] The influence index updating module establishes an OCSVM model, is used to receive the first potential customer group, calculate the anomaly scores of all the first potential customers in the OCSVM model, and correct the preset influence index of the customer group to which these first potential customers belong according to the anomaly scores of the first potential customers;

[0041] The secondary prediction module is used to re-input the updated second data set into the logistic regression model to output a second potential customer group;

[0042] The active registration guidance module sends guidance information to the second potential customer group through a guidance channel.

[0043] In a preferred embodiment of the present invention, the impact index updating module further comprises: an abnormal score mean value calculation unit, an impact index adjustment factor calculation unit, a new impact index calculation unit and an allocation unit;

[0044] The anomaly score mean calculation unit is used to calculate the average anomaly score of each customer group , and the average anomaly score of all customer groups , is the total number of customers;

[0045] The influence index adjustment factor calculation unit further calculates the customer group influence index adjustment factor according to the output of the abnormal score mean calculation unit. , is the preset influence index of customer group j;

[0046] The new influence index calculation unit adds the influence index adjustment factor to the preset influence index to obtain the unnormalized new influence index of each customer group, and then calculates the reciprocal of the sum of all the unnormalized new influence indexes. , further calculate the normalized new impact index of each customer group ;

[0047] The allocation unit converts the normalized new impact index of each customer group Update to each customer group in the second dataset.

[0048] In a preferred embodiment of the present invention, the guiding channels include telephone marketing, email marketing, social media, mobile application push and website pop-up windows.

[0049] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0050] The present invention provides a method for mining potential customers and actively registering and guiding customers based on big data analysis. First, a large amount of customer data is subjected to similarity analysis to obtain the relationship between different customers. Based on the similarity analysis results, several customer groups are divided, and different customer groups are assigned different influence indexes. Then, a logistic regression model is input to predict and identify potential customers in the customer data. After identifying the potential customers, a personalized guidance strategy is designed according to the characteristics of each customer group, thereby improving the recognition rate of potential customers and ensuring the pertinence and efficiency of marketing activities. Compared with the prior art, the present invention effectively solves the problem of category imbalance in big data analysis and significantly improves the accuracy and generalization ability of data processing.

[0051] The present invention first collects a large amount of customer data, and then implements similarity analysis, based on which the customer data is accurately divided into different customer groups, and the corresponding influence index is preset for each customer group category. The second data set constructed in this way is a collection of several customer groups and corresponding preset influence indexes. The preset influence index is specifically to assign a higher influence index to a minority category of customer groups, so that the several customer groups in the second data set are balanced in size when used as input into the logistic regression model, avoiding the problem of data deviation during the analysis of the logistic regression model, and improving the model's ability to identify a minority of customer groups, thereby more effectively mining potential customers.

[0052] The present invention combines the logistic regression model and the OCSVM model, outputs the first potential customer by using the logistic regression model, calculates the anomaly score of the first potential customer by using the trained OCSVM model, reversely updates the preset influence index of each customer group in the second data set according to the calculated anomaly score, and then re-inputs it into the logistic regression model to predict and output the second potential customer. Through the above steps, data support can be provided for the setting of the influence index of each customer group in the second data set, a higher influence index can be set for customer groups that are more similar to purchased customers, and the volume of several customer groups in the second data set can be further balanced when they are input into the logistic regression model, avoiding the problem of data deviation during the analysis of the logistic regression model, and improving the accuracy and generalization ability of the logistic regression model. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 It is a flow chart of a method for potential customer mining and active registration guidance based on big data analysis according to a preferred embodiment of the present invention;

[0055] Figure 2 is a flow chart of a similarity analysis method according to a preferred embodiment of the present invention;

[0056] Figure 3 is a flow chart of a method for calculating anomaly scores of first potential customers according to a preferred embodiment of the present invention;

[0057] Figure 4 It is a schematic diagram of a potential customer mining and active registration guidance system based on big data analysis according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0060] like Figure 1 As shown, the present invention provides a method for potential customer mining and active registration guidance based on big data analysis, comprising the following steps:

[0061] S1. Collect a large amount of customer data, including customer data of products that have been purchased and those that have not been purchased, and clean and standardize the collected customer data to obtain a first data set;

[0062] S2. Performing similarity analysis on the customers in the first data set, identifying customer similarities between a number of customers in the first data set, and dividing the customers in the first data set into different customer groups based on the similarity analysis results, wherein a number of customers in the same customer group have similar purchasing behaviors and preferences;

[0063] S3, presetting an influence index for each customer group to obtain a second data set;

[0064] S4. Input the second data set into the logistic regression model, and output a first potential customer group, where the first potential customers are customers who have purchase intentions but have not yet made a purchase. The logistic regression model marks these customers as abnormal, i.e., a group different from the group of customers who have made a purchase;

[0065] S5, inputting the first potential customer group into the OCSVM model, calculating the anomaly scores of all first potential customers, modifying the preset influence index of the customer group to which these first potential customers belong according to the anomaly scores of the first potential customers, re-inputting the updated second data set into the logistic regression model, and outputting the second potential customers;

[0066] S6. Sending guidance information to the second potential customer group through the guidance channel, specifically guiding the second potential customer group to register and trade through customized products or services.

[0067] The present invention provides a potential customer mining and active registration guidance method based on big data analysis. By combining similarity analysis, grouping, weighted logistic regression and other steps, potential customers can be identified more comprehensively and accurately, thereby improving the success rate and efficiency of marketing activities.

[0068] The products in step S1 include: financial products such as funds, trusts, bonds and stocks, as well as commercial, medical, life and accident insurance products.

[0069] The data sources in step S1 include, but are not limited to, online trading platforms, customer relationship management systems (CRMs), social media, website traffic analysis tools, etc. The acquisition of such data involves database queries, API calls, web crawler technology, etc.

[0070] Data cleaning includes: removing duplicate records, processing missing values, identifying and processing outliers, etc. These are existing technologies and will not be elaborated here.

[0071] Standardization processing includes normalizing numerical features to eliminate the influence of dimension, normalizing categorical features to unify data format, and integrating data into the first data set for subsequent processing and analysis.

[0072] It is worth noting that the customer data in step S1 includes but is not limited to: basic customer information (name, gender, age, contact information and address, etc.), financial status (income level, total assets, liabilities, credit score, purchasing power score, etc.), purchase history (types, quantity, price, purchase time, etc. of goods purchased), browsing behavior (browsed goods, browsing time, browsing depth, etc.), user feedback (evaluation, rating, complaint and other feedback information on goods or services), social media behavior (customer behavior on social media, such as likes, shares, comments, etc.), user registration information (customer registration time, registration channel, preferences filled in during registration, etc.), user interaction (records of interaction between customers and the platform, such as marketing activities participated in, mailing lists subscribed, etc.), geographic location information (customer geographic location information, which may be obtained through IP address, GPS, etc.), etc.

[0073] The present invention constructs a clean and standardized first data set through the data collection in step S1 and the data preprocessing step, laying a solid foundation for subsequent similarity analysis, customer group segmentation and model training.

[0074] like Figure 2 As shown, the similarity analysis in step S2 includes the following steps:

[0075] S21. Select features related to customer purchasing behavior and preferences from the first data set, including but not limited to: purchase history, browsing behavior, user feedback, transaction and payment information, user interaction, and device and application usage information.

[0076] S22: Perform similarity calculation on the features selected in step S21, and calculate the similarity score between each pair of customers.

[0077] It is worth noting that for non-numeric data, non-numeric data can be converted into numerical data through data conversion methods such as binary encoding, one-hot encoding or label encoding.

[0078] Since different features have different dimensions and value ranges, the features need to be standardized before similarity calculation to ensure that the influence index of each feature on the distance calculation is the same. In this embodiment, the Z score standardization method is preferably used: ,in is the original data, is the mean of the feature, is the standard deviation.

[0079] The standardized data is obtained by calculation, and then the Euclidean distance between each pair of customers is calculated: ,in and are the normalized feature vectors of the two customers, yes The characteristic value of yes The characteristic value of .

[0080] S23. Construct the calculated similarity scores into a similarity matrix, where each element in the matrix represents the similarity between a pair of customers.

[0081] The elements on the diagonal of the similarity matrix are the Euclidean distances between each pair of customers. The smaller the Euclidean distance value, the higher the similarity between the two customers.

[0082] In step S2, based on the similarity matrix obtained in step S23, a clustering algorithm, preferably K-means, is used to divide the customers into different customer groups, and the similarity between customers in each customer group exceeds a set similarity threshold.

[0083] The similarity threshold in this embodiment is 0.6, 0.8, 1.0, 1.1, etc., and the similarity threshold is set according to actual needs. When the similarity between a pair of customers is greater than the similarity threshold, it means that the two customers have similar purchasing behaviors and preferences and are classified into the same customer group.

[0084] The present invention first collects a large amount of customer data, and then implements similarity analysis, based on which the customer data is accurately divided into different customer groups, and the corresponding influence index is preset for each customer group category. The second data set constructed in this way is a collection of several customer groups and corresponding preset influence indexes. The preset influence index is specifically to assign a higher influence index to a minority category of customer groups, so that the several customer groups in the second data set are balanced in size when used as input into the logistic regression model, avoiding the problem of data deviation during the analysis of the logistic regression model, and improving the model's ability to identify a minority of customer groups, thereby more effectively mining potential customers.

[0085] In step S3, the method of presetting the influence index is specifically as follows: presetting the influence index for each customer group based on prior experience, that is, by reviewing and analyzing historical customer data, identifying which customer groups have historically had higher purchase conversion rates, and which factors have a significant impact on the purchasing behavior of these customer groups, and based on the collected prior experience, presetting an initial influence index for each customer group, which reflects the importance of the customer group in the logistic regression model.

[0086] The construction of the logistic regression model in this embodiment includes the following steps:

[0087] A1. Divide the second data set into a training set and a test set. The training set is used to build the model, and the test set is used to evaluate the prediction performance of the logistic regression model.

[0088] It is worth noting that if the amount of customer information in the second data set is small, the accuracy and generalization ability of the logistic regression model trained using it will be greatly limited. In this embodiment, data synthesis is used, preferably a variational autoencoder is used to expand the amount of customer information in the second data set for training the logistic regression model, thereby improving the accuracy and generalization ability of the logistic regression model.

[0089] A2. Use the training set data to build a logistic regression model, and adjust the sample influence index in the logistic regression model according to the preset influence index of the customer group in the second data set. The logistic regression model will predict whether the customer is a potential customer, that is, whether he has the intention to buy but has not yet made a purchase.

[0090] A3. Use the training set with influence index adjustment to train the logistic regression model. During the training process, the model will learn the relationship between the characteristics of the customer information in the training set and the target variable (i.e., whether the customer is a potential customer).

[0091] A4. Use the test set data to evaluate the trained logistic regression model. The evaluation indicators include accuracy, recall, F1 score, etc.

[0092] A5. Adjust and optimize the model based on the results of model evaluation, including adjusting feature selection, influence index settings or model parameters, to improve the prediction accuracy of the model.

[0093] By constructing a logistic regression model, we can effectively identify potential customers who have purchasing intentions but have not yet purchased financial products from a large amount of customer data, and provide financial institutions with more targeted financial products or services to guide them to register and trade, thereby improving the success rate of marketing activities and customer satisfaction.

[0094] It is worth noting that the output of the logistic regression model is a probability value, which indicates the probability that a non-purchasing customer is a potential customer. In practical applications, a probability threshold is set according to the demand, customers with probability values ​​greater than the threshold are marked as potential customers, and the list of potential customers predicted by the logistic regression model is output.

[0095] like Figure 3 As shown, the calculation of the anomaly score for the first potential customer in step S5 includes the following steps:

[0096] S51. Use the python framework to build the OCSVM model. Use the OCSVM algorithm to separate the zero data point in the feature space from all data points, maximize the distance from the zero point to the separation hyperplane, and create a decision hyperplane. The OCSVM model is an existing technology and will not be described in detail here.

[0097] S52, the purchased customer data in the second data set is divided into a training set and a test set to train the OCSVM model. After training, a decision boundary is obtained in the OCSVM model, which includes the purchased customer data in the second data set and excludes abnormal points (i.e., non-purchased customer data, which does not exist in the training set but may exist in actual applications). The number of customers in the training set is more than twice that of the test set.

[0098] S53. Input the first potential customer into the trained OCSVM model, and calculate the anomaly score of each potential customer. The anomaly score is the distance from the data point of the potential customer mapped to the decision hyperplane to the decision boundary. The higher the anomaly score, the greater the difference in characteristics between the first potential customer and the group of purchased customers.

[0099] S54. According to the abnormal score of the potential customer, the preset influence index of each customer group in the second data set is reversely updated and re-input into the logistic regression model to output the second potential customer.

[0100] Updates to the preset impact index are as follows:

[0101] Divide the first potential customer group into original customer groups and calculate the average anomaly score for each customer group ;

[0102] Calculate the average anomaly score for all customer groups , is the total number of customers;

[0103] Calculate the customer impact index adjustment factor , is the preset influence index of customer group j;

[0104] Add the impact index adjustment factor to the preset impact index to obtain the unnormalized new impact index for each customer group, and then calculate the reciprocal of the sum of all unnormalized new impact indices. ;

[0105] Calculate the normalized new impact index for each customer group ;

[0106] The normalized new impact index of each customer group Update to each customer group in the second dataset.

[0107] In this embodiment, when there is no potential customer in a certain customer group in the second data set, the lowest influence index among the preset influence indexes is uniformly allocated to the customer group without potential customers.

[0108] The present invention combines the logistic regression model and the OCSVM model, outputs the first potential customer by using the logistic regression model, calculates the anomaly score of the first potential customer by using the trained OCSVM model, reversely updates the preset influence index of each customer group in the second data set according to the calculated anomaly score, and then re-inputs it into the logistic regression model to predict and output the second potential customer. Through the above steps, data support can be provided for the setting of the influence index of each customer group in the second data set, a higher influence index can be set for customer groups that are more similar to purchased customers, and the volume of several customer groups in the second data set can be further balanced when they are input into the logistic regression model, avoiding the problem of data deviation during the analysis of the logistic regression model, and improving the accuracy and generalization ability of the logistic regression model.

[0109] In step S6, guidance information is sent to the second potential customer group through the guidance channel, specifically guiding the second potential customer group to register and trade through customized products or services.

[0110] According to the purchase behavior, preference, geographic location and other information of the second potential customer output in step S5, the second potential customer is subdivided into different groups, and a targeted guidance strategy is formulated, including providing personalized content, discounts, services, etc. Guidance channels include email marketing, social media, mobile application push and website pop-up windows, etc.

[0111] In a method for prospecting and actively registering financial products, there are two customer groups with different behavioral preferences in the information output by the logistic regression model, namely the high-frequency trading group and the long-term investment group. Different guidance strategies are implemented for these two groups:

[0112] For high-frequency traders who tend to trade frequently and are sensitive to market dynamics and investment opportunities, the guidance strategy implemented is to provide real-time market information, trading strategy guides, high-frequency trading discounts, etc. For example, through email marketing, daily market updates and trading tips are sent, and limited-time trading commission reduction discounts are also provided.

[0113] The long-term investment group focuses on long-term returns and prefers stable and low-risk investment products. The guiding strategy implemented is to introduce the benefits of long-term investment, provide detailed information on stable return products, and exclusive offers for long-term investors, such as: publishing long-term investment case analysis and product recommendations through social media platforms, and launching special financial management plans for long-term investors.

[0114] Furthermore, for customer groups with different geographical locations:

[0115] For local customer groups located in the city where the company is located or in nearby areas, the guidance strategy implemented is to provide offline event invitations, face-to-face consultation opportunities, and investment products with local characteristics, such as: pushing local investment lecture information through mobile applications, and providing appointment services for face-to-face consultations.

[0116] For the out-of-town customer group that is far away from the company's location and distributed across the country, the guidance strategy implemented is to provide remote consulting services, online investment education courses, and convenient online trading experience. For example: through website pop-ups, the introduction and appointment method of the remote consulting expert are displayed, and the link to the online investment course is pushed.

[0117] Furthermore, different levels of discounts can be provided based on the customer's historical investment amount, trading frequency, etc. For example, exclusive discount coupons for large investors can be sent via email, or points can be given based on trading frequency.

[0118] The present invention provides a method for mining potential customers and actively registering and guiding customers based on big data analysis. First, a large amount of customer data is subjected to similarity analysis to obtain the relationship between different customers. Based on the similarity analysis results, several customer groups are divided, and different customer groups are assigned different influence indexes. Then, a logistic regression model is input to predict and identify potential customers in the customer data. After identifying the potential customers, a personalized guidance strategy is designed according to the characteristics of each customer group, thereby improving the recognition rate of potential customers and ensuring the pertinence and efficiency of marketing activities. Compared with the prior art, the present invention effectively solves the problem of category imbalance in big data analysis and significantly improves the accuracy and generalization ability of data processing.

[0119] like Figure 4 As shown, the present invention also provides a potential customer mining and active registration guidance system based on big data analysis, including: a data collection module, a data preprocessing module, a similarity analysis module, a grouping module, an influence index preset module, a primary prediction module, an influence index update module, a secondary prediction module and an active registration guidance module; the data collection module collects a large amount of customer data from multiple channels; the data preprocessing module is used to perform data cleaning and standardization processing on the original customer data collected by the data collection module, and output a first data set; the similarity analysis module is used to perform similarity analysis on the customer information in the first data set and calculate the similarity between each pair of customers; the grouping module divides the customers in the first data set into different customer groups according to the similarity analysis results, and several customers in the same customer group have similar Purchasing behavior and preference; an influence index preset module, which presets an influence index for each customer group and combines it into a second data set as input to the prediction module; a primary prediction module, which establishes a logistic regression model, inputs the second data set, analyzes and predicts whether the customers in the second data set are potential customers, and outputs the first potential customer group; an influence index update module, which establishes an OCSVM model for receiving the first potential customer group, calculates the anomaly scores of all first potential customers in the OCSVM model, and corrects the preset influence index of the customer group to which these first potential customers belong based on the anomaly scores of the first potential customers; a secondary prediction module, which is used to re-input the updated second data set into the logistic regression model and output the second potential customer group; an active registration guidance module, which sends guidance information to the second potential customer group through the guidance channel.

[0120] In summary, the present invention provides a potential customer mining and active registration guidance system based on big data analysis. First, a large amount of customer data is subjected to similarity analysis to obtain the relationship between different customers. Based on the similarity analysis results, several customer groups are divided, and different customer groups are assigned different influence indexes. Then, a logistic regression model is input to predict and identify potential customers in customer data. After identifying potential customers, a personalized guidance strategy is designed according to the characteristics of each customer group, thereby improving the recognition rate of potential customers and ensuring the pertinence and efficiency of marketing activities. Compared with the prior art, the present invention effectively solves the problem of category imbalance in big data analysis and significantly improves the accuracy and generalization ability of data processing.

[0121] The influence index update module in this embodiment also includes: an abnormal score mean calculation unit, an influence index adjustment factor calculation unit, a new influence index calculation unit and an allocation unit; an abnormal score mean calculation unit is used to calculate the average abnormal score of each customer group. , and the average anomaly score of all customer groups , is the total number of customers; the impact index adjustment factor calculation unit further calculates the customer impact index adjustment factor based on the output of the abnormal score mean calculation unit , is the preset influence index of customer group j; the new influence index calculation unit adds the influence index adjustment factor to the preset influence index to obtain the unnormalized new influence index of each customer group, and then calculates the reciprocal of the sum of all unnormalized new influence indices , further calculate the normalized new impact index of each customer group ; Allocation unit, normalized new impact index of each customer group Update to each customer group in the second dataset.

[0122] The setting of the influence index update module of the present invention combines the logistic regression model and the OCSVM model, outputs the first potential customer by using the logistic regression model, calculates the anomaly score of the first potential customer by using the trained OCSVM model, reversely updates the preset influence index of each customer group in the second data set according to the calculated anomaly score, and then re-inputs it into the logistic regression model to predict and output the second potential customer. Through the above steps, data support can be provided for the setting of the influence index of each customer group in the second data set, a higher influence index can be set for customer groups that are more similar to purchased customers, and the volume of several customer groups in the second data set can be further balanced when they are input into the logistic regression model, avoiding the problem of data deviation during the analysis of the logistic regression model, and improving the accuracy and generalization ability of the logistic regression model.

[0123] The guidance channels in this embodiment include telephone marketing, email marketing, social media, mobile application push and website pop-up windows, etc.

[0124] The above functions described in the present invention can be implemented in hardware, software, firmware or any combination of the three. If implemented in software, these functions can be stored on a computer-readable medium, or transmitted on a computer-readable medium in the form of one or more instructions or codes. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one place to another. The storage medium can be any available medium that can be accessed by any general or special computer. For example, such computer-readable media can include but are not limited to RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store instructions or data structures and other program codes in the form that can be read by general or special computers, or general or special processors. In addition, any connection may be appropriately defined as a computer-readable medium, for example, if the software is transmitted from a website, server or other remote resource via a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wireless methods such as infrared, radio and microwave. Disks and discs are also included in the definition of computer-readable media. Disks generally reproduce data magnetically, while discs generally reproduce data optically using lasers. Combinations of the above may also be included in computer-readable media.

[0125] The above is based on the ideal embodiment of the present invention. Through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. A method for mining potential customers and actively guiding registration based on big data analysis, characterized in that: The following steps are involved: S1. Collect a large amount of customer data, including customer data of products that have been purchased and those that have not been purchased, and clean and standardize the collected customer data to obtain a first data set; wherein the customer data includes: purchase behavior, preference and geographic location; S2. Performing similarity analysis on the customers in the first data set, identifying customer similarities between a number of customers in the first data set, and dividing the customers in the first data set into different customer groups based on the similarity analysis results, wherein a number of customers in the same customer group have similar purchasing behaviors and preferences; The similarity analysis in S2 includes the following steps: S21, selecting features related to customer purchasing behavior and preferences from the first data set; S22, performing similarity calculation on the features selected in step S21, and calculating the similarity score between each pair of customers; S23, constructing a similarity matrix using the calculated similarity scores, where each element in the matrix represents the similarity between a pair of customers; In S22, the selected features are standardized before similarity calculation, using the Z-score standardization method: ,in is the original data, is the mean of the feature, is the standard deviation; The standardized data is obtained by calculation, and then the Euclidean distance between each pair of customers is calculated: ,in and are the normalized feature vectors of the two customers, yes The characteristic value of yes The characteristic value of S3, presetting an influence index for each customer group to obtain a second data set; S4. Input the second data set into the logistic regression model, and output a first potential customer group, where the first potential customers are customers who have purchase intentions but have not yet made a purchase. The logistic regression model marks these customers as abnormal. S5, inputting the first potential customer group into the OCSVM model, calculating the anomaly scores of all the first potential customers, modifying the preset influence index of the customer group to which these first potential customers belong according to the anomaly scores of the first potential customers, re-inputting the updated second data set into the logistic regression model, and outputting the second potential customer group; The calculation of the anomaly score for the first potential customer in S5 includes the following steps: S51. Use the python framework to build the OCSVM model; S52, dividing the purchased customer data in the second data set into a training set and a test set to train the OCSVM model. After the training, a decision boundary is obtained in the OCSVM model, and the decision boundary includes the purchased customer data in the second data set and excludes the abnormal points; S53, inputting the first potential customer into the trained OCSVM model, and calculating the anomaly score of each potential customer, where the anomaly score is the distance from the data point of the potential customer mapped to the decision hyperplane to the decision boundary. The higher the anomaly score, the greater the difference in characteristics between the first potential customer and the group of purchased customers. The updating of the preset impact index is specifically as follows: Divide the first potential customer group into original customer groups and calculate the average anomaly score for each customer group ; Calculate the average anomaly score for all customer groups , is the total number of customers; Calculate the customer impact index adjustment factor , is the preset influence index of customer group j; Add the impact index adjustment factor to the preset impact index to obtain the unnormalized new impact index for each customer group, and then calculate the reciprocal of the sum of all unnormalized new impact indices. ; Calculate the normalized new impact index for each customer group ; The normalized new impact index of each customer group Update to each customer group in the second data set; S6. Send guidance information to the second potential customer group through the guidance channel, and guide the second potential customer group to register and trade through customized products or services.

2. The method for potential customer mining and active registration guidance based on big data analysis according to claim 1 is characterized by: The non-numeric data in the feature is converted into numeric data through binary encoding, one-hot encoding or label encoding method.

3. The method for potential customer mining and active registration guidance based on big data analysis according to claim 1 is characterized by: In S6, the second potential customers are segmented into different groups according to the purchase behavior, preference and geographic location information of the second potential customers outputted in S5, and targeted guidance strategies are formulated, including providing personalized content, discounts and services.

4. A system for mining potential customers and actively registering and guiding customers based on big data analysis, using the method for mining potential customers and actively registering and guiding customers based on big data analysis according to any one of claims 1 to 3, characterized in that: include: Data collection module, data preprocessing module, similarity analysis module, grouping module, influence index preset module, primary prediction module, influence index update module, secondary prediction module and active registration guidance module; The data collection module collects a large amount of customer data from multiple channels; The data preprocessing module is used to perform data cleaning and standardization processing on the original customer data collected by the data collection module, and output a first data set; The similarity analysis module is used to perform similarity analysis on the customer information in the first data set and calculate the similarity between each pair of customers; The grouping module divides the customers in the first data set into different customer groups according to the similarity analysis result, and several customers in the same customer group have similar purchasing behaviors and preferences; The influence index preset module presets an influence index for each customer group and combines them into a second data set as input to the prediction module; The primary prediction module establishes a logistic regression model, inputs the second data set, analyzes and predicts whether the customers in the second data set are potential customers, and outputs a first potential customer group; The influence index updating module establishes an OCSVM model, is used to receive the first potential customer group, calculate the anomaly scores of all the first potential customers in the OCSVM model, and correct the preset influence index of the customer group to which these first potential customers belong according to the anomaly scores of the first potential customers; The impact index update module further includes: an abnormal score mean value calculation unit, an impact index adjustment factor calculation unit, a new impact index calculation unit and an allocation unit; The anomaly score mean calculation unit is used to calculate the average anomaly score of each customer group , and the average anomaly score of all customer groups , is the total number of customers; The influence index adjustment factor calculation unit further calculates the customer group influence index adjustment factor according to the output of the abnormal score mean calculation unit. , is the preset influence index of customer group j; The new influence index calculation unit adds the influence index adjustment factor to the preset influence index to obtain the unnormalized new influence index of each customer group, and then calculates the reciprocal of the sum of all the unnormalized new influence indexes. , further calculate the normalized new impact index of each customer group ; The allocation unit converts the normalized new impact index of each customer group Update to each customer group in the second data set; The secondary prediction module is used to re-input the updated second data set into the logistic regression model to output a second potential customer group; The active registration guidance module sends guidance information to the second potential customer group through a guidance channel.

5. The potential customer mining and active registration guidance system based on big data analysis according to claim 4 is characterized by: The said guidance channels include telephone marketing, email marketing, social media, mobile application push and website pop-ups.

Citation Information

Patent Citations

  • Evaluation and analysis method of agricultural product business model and information digitization system thereof

    CN118278606A

  • Information data management method based on process mining technology

    CN118279067A