A customer data background management system and its use method

Abstract

The present invention discloses a customer data background management system and a use method thereof, which relates to the technical field of customer data background management, and comprises a customer authentication module, a customer data collection module, a customer data storage module, a customer data analysis and processing module, and a customer data visualization module. The customer data background management system ensures safe login through password verification and SMS authentication; the data collection module reads customer data from three data sources and updates it regularly; the data storage module manages data update strategies by using smart contracts, encrypts and stores the data in a MySQL database, and performs regular backups; feature selection is performed using a random forest importance analysis method and a Lasso method based on the collected customer data, and customer credit scores are predicted based on the selected customer features, so that enterprises can more accurately evaluate customer credit; intuitive analysis results are presented through a visualization module, and trends in customer prediction results are quickly understood, thereby accurately formulating operation strategies.

Description

Technical Field

[0001] The present invention relates to the technical field of customer data background management, and in particular to the technical field of a customer data background management system and a method for using the same. Background Art

[0002] With the rapid development of information technology, customer data management has become an indispensable part of enterprise operations. Effective customer data management not only improves the market competitiveness of enterprises, but also enhances customer satisfaction and loyalty. However, existing customer data management systems usually only focus on data collection and storage, and lack the ability to conduct in-depth analysis and visualization of data, which limits enterprises from extracting valuable information from massive data.

[0003] In the Chinese invention application with reapplication publication number CN112883096A, a customer data management system is disclosed, in which an acquisition module is used to acquire original business data from a data warehouse, and allocate it to corresponding business data sub-domains according to the business type of the original business data; wherein the data type of the original business data includes internal characteristic tags, mining model tags and various market data tags; the aggregation module is used to mark and aggregate the data according to the business data sub-domains to the business data general domain; the transmission module is used to send the original business data with business tags in the business data general domain to the customer data general domain; the allocation module is used to generate original customer data based on the original business data with business tags in the customer data general domain, and allocate it to corresponding customer data business layer sub-domains according to the customer type of the original customer data; the convergence module is used to mark and aggregate the data according to the customer data business layer sub-domains to the customer data business layer general domain.

[0004] In the above invention application, the original business data is obtained through the data warehouse and allocated to the corresponding business data domain according to the business type of the original business data. However, there is a lack of security measures in the data transmission process, as well as a lack of user authority management and access control mechanism, and insufficient attention to data privacy and compliance. In addition, the data is not pre-processed and the relevant customer data is not predicted and visualized through the model. This has great limitations in practical applications, resulting in the inability to meet the actual use requirements of the customer data background management system.

[0005] To this end, the present invention provides a customer data background management system and a method for using the same. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In view of the deficiencies in the prior art, the present invention provides a customer data background management system and a method for using the same. The present invention introduces passwords and SMS verification codes for identity authentication and authorization, obtains customer information from forms, financial institutions and online questionnaires through real-time event-driven and periodic polling, uses Ethereum blockchain and smart contracts to manage data updates, utilizes the PoS consensus mechanism to determine data consistency, saves encrypted data to a MySQL database, performs regular backups, preprocesses, performs feature selection and SVR modeling on stored data, predicts customer credit scores, and provides intuitive data visualization functions to comprehensively improve the security, availability and data analysis capabilities of the system, thereby solving the technical problems recorded in the background technology.

[0008] (II) Technical solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: a customer data background management system, comprising:

[0010] Customer authentication module, used for customer identity authentication and authorization, with multi-factor authentication, including password verification and SMS verification code;

[0011] The customer data collection module updates customer data through real-time event-driven mechanisms and periodic polling, and uses data collection forms to collect key customer information, including customer number, age, income level, years of work experience, and family size; obtains credit scores, asset values, property values, health expenses, and insurance amounts through financial institutions; and obtains customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values, and loan interest rates through online questionnaires;

[0012] The customer data storage module is used to receive data from the data collection module. When the customer data is updated, it uses the Ethereum blockchain platform and smart contracts to manage data updates. It then uses the PoS (Proof of Stake) consensus mechanism to determine the consistency of the updated data, encrypt the data, and save it in the MySQL database, implementing a regular data backup strategy.

[0013] The customer data analysis and processing module is used to analyze and process the stored customer data information, including data preprocessing, feature selection and SVR model building, and predict customer credit scores;

[0014] The data visualization module provides intuitive data results and generates visual reports, charts and data dashboards.

[0015] Furthermore, customer data is obtained, including obtaining customer key information through data collection forms, financial institutions and online questionnaires. The collected customer data includes: customer number, age, income level, years of work, family size, credit score, asset value, property value, health expenses, insurance amount, account balance, repayment record, purchase frequency, deposit amount, credit card overdraft limit, vehicle value and loan interest rate, which are stored in the customer data storage module.

[0016] Furthermore, when customer data is updated, the Ethereum blockchain platform is used to manage data updates using smart contracts, and the PoS (Proof of Stake) consensus mechanism is used to determine the consistency of the updated data, encrypt the data, save it in the MySQL database, and implement a regular data backup strategy.

[0017] Furthermore, the customer data stored in the customer data storage module is preprocessed and feature selected, and the data features with high correlation with the credit score are selected using the random forest importance analysis method. The feature vector is , where each For one of the features, k ranges from 1-16.

[0018] Furthermore, the importance of the feature variable (customer information) and the target variable (customer credit score) is analyzed using the random forest importance analysis method to obtain the feature importance vector , use the Lasso method to analyze the correlation between the feature variable (customer information) and the target variable (customer need score), and obtain the regression coefficient of each feature through the Lasso method ,calculate The absolute value of the coefficient =( ,...., )}.

[0019] Furthermore, the importance score of the random forest and the absolute value of the regression coefficient of the Lasso method are calculated and weighted averaged to obtain the comprehensive value of customer characteristics. The formula is as follows:

[0020]

[0021] Among them, p is 0.5, k is the index of the kth feature, and then calculate the 80th percentile threshold T of the feature comprehensive value, and select features as the final features.

[0022] Furthermore, using the customer data obtained by feature selection, the customer credit score is predicted by the SVR model. The prediction function formula is as follows: For a given training data set where x i are input features, is the corresponding target value. The goal of the SVR model is to find a prediction function , the formula is as follows:

[0023]

[0024] Where w is the weight vector, T is the transposed sign, b is the bias coefficient, Yes Input The feature map of The result after mapping to high-dimensional space, kernel function Provides a measure of the similarity or distance between two samples, where represents the input feature vector of the i-th sample in the training data set, represents the input feature vector of the jth sample in the training data set, and N is the number of training data samples.

[0025] Furthermore, the optimization problem is established. SVR optimizes the model by minimizing the loss function and the regularization term to determine the optimal w and b. SVR optimizes the following objective problems as shown below:

[0026]

[0027] in, is a regularization term used to prevent overfitting and ensure the simplicity of the model. C is a penalty factor that weighs the impact of regularization and loss function. and are slack variables for the upper and lower bounds, and the constraints are:

[0028]

[0029] in, is the insensitive loss factor.

[0030] Furthermore, the Lagrange multiplier is introduced to make it easier to solve. The dual problem is as follows:

[0031]

[0032] The constraints are:

[0033]

[0034] Among them, the introduction of two sample indexes i and j is used to construct the optimization problem. and is the Lagrange multiplier of sample i and is the Lagrange multiplier of sample j, and the kernel function is used to describe and optimize the relationship between samples in the training data set.

[0035] Furthermore, the dual problem is solved by optimizing the dual problem above, and the Lagrange multiplier is obtained. Then w and b are calculated based on the multiplier. The formula is as follows:

[0036]

[0037]

[0038] Through the above process, w and b are calculated, and the test set data is used to evaluate the performance of the model. The evaluation indicators used are determination coefficient, mean square error, root mean square error, and mean absolute error. The determination coefficient reflects the proportion of the variance of the predicted data explained by the model. The closer the value is to 1, the stronger the explanatory ability is. The mean square error and root mean square error represent the average of the square difference between the predicted value and the true value and its square root, respectively. The smaller the two are, the higher the prediction accuracy is. The mean absolute error measures the average absolute value of the prediction error. The smaller the value is, the higher the prediction accuracy of the model is. These indicators evaluate the accuracy, explanatory power and error distribution of the model. The specific formulas are as follows:

[0039]

[0040]

[0041]

[0042]

[0043] in, is the actual value, is the predicted value, is the average of the actual values. The comprehensive evaluation index D is used to understand the customer credit score. The formula is as follows:

[0044]

[0045] Among them, when w is 0.25, the comprehensive evaluation index 1≤ When <10, the fitting effect is better and the model predicts the credit score more accurately; 10≤ When D <50, the fitting effect is average and the predicted credit score result is average; when D ≥50, the model predicts the credit score result poorly. When the SVR model predicts the customer's credit score, if the comprehensive evaluation index is average or poor, it means that the model prediction is inaccurate and the model's hyperparameters need to be adjusted to improve the model's accuracy.

[0046] Furthermore, the data visualization module provides intuitive data results and generates visual reports, charts and data dashboards.

[0047] A method for backend management of customer data includes the following steps:

[0048] Used for customer identity authentication and authorization, where multi-factor authentication includes password verification and SMS verification code;

[0049] Used to obtain customer information, update customer data through real-time event-driven mechanisms and regular polling, collect customer numbers, ages, income levels, years of work experience, and family size through data collection forms; obtain credit scores, asset values, property values, health expenses, and insurance amounts through financial institutions; obtain customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values, and loan interest rates through online questionnaires;

[0050] It is used to receive data from the data collection module. When the customer data is updated, it uses the Ethereum blockchain platform and smart contracts to manage data updates. It then uses the PoS (Proof of Stake) consensus mechanism to determine the consistency of the updated data, encrypt the data, save it in the MySQL database, and implement a regular data backup strategy.

[0051] Used to analyze and process stored customer data information, including data preprocessing, feature selection, and support vector regression (SVR) model building to predict customer credit scores;

[0052] Used to provide intuitive data results and generate visual reports, charts and data dashboards.

[0053] (III) Beneficial effects

[0054] The present invention provides a customer data background management system and a method for using the same, which have the following beneficial effects:

[0055] 1. The customer authentication module adopts multi-factor authentication, including password verification and SMS verification code, to ensure the security and accuracy of customer identity authentication.

[0056] 2. The customer data collection module obtains comprehensive key customer information through three channels: data collection forms, financial institutions and online questionnaires. It uses real-time event-driven mechanisms and regular polling to update customer data to ensure the comprehensiveness and timeliness of key customer information.

[0057] 3. The customer data storage module uses the MySQL database to receive data from the data collection module. When the customer data is updated, the Ethereum blockchain platform is used to manage data updates using smart contracts. The PoS (Proof of Stake) consensus mechanism is then used to determine the consistency of the updated data, encrypt the data, and save it in the MySQL database. A regular data backup strategy is implemented to improve data security and durability.

[0058] 4. The customer data analysis and processing module conducts in-depth analysis and processing of customer data through data preprocessing, feature selection and SVR model building, accurately predicts customer credit scores, and makes more accurate risk assessments and decisions.

[0059] 5. The data visualization module provides intuitive data result display, generates visual reports, charts and data dashboards, and can quickly understand data trends, discover patterns, and conduct decision analysis to optimize customer service and business operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0061] Figure 1 A schematic diagram of the structure of a customer data background management system of the present invention;

[0062] Figure 2 The present invention is a flowchart of a method for backstage management of customer data. DETAILED DESCRIPTION

[0063] 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.

[0064] See also Figure 1 The present invention provides a customer data background management system, comprising:

[0065] Customer authentication module, customer data collection module, customer data storage module, customer data analysis and processing module, customer data visualization module;

[0066] The customer authentication module confirms the customer's identity by verifying the correctness of the username and password, and has enabled two-factor authentication. The customer is also required to enter a verification code sent to their mobile phone or email to successfully log in. The module also monitors login behavior in real time, detects abnormal activities (such as logging in from a different location, multiple failed login attempts in a short period of time), triggers an alarm, and requires the user to perform additional identity verification.

[0067] The customer data collection module obtains data through data collection forms, including customer number, age, income level, years of work and family population; obtains credit score, asset value, property value, health expenditure and insurance amount information through financial institutions; obtains customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values ​​and loan interest rates through online questionnaires, and updates customer data using real-time event-driven mechanisms and regular polling.

[0068] The collected customer information is stored in the MySQL database. When the customer data is updated, the Ethereum blockchain platform is used to manage the data update using smart contracts. The PoS (Proof of Stake) consensus mechanism is then used to determine the consistency of the updated data. The data is encrypted and saved in the MySQL database, and a regular data backup strategy is implemented.

[0069] Customer data analysis and processing module, the customer data background management system predicts the customer's credit score through other customer information data. The data obtained by the data collection module is saved in the MySQL database, and the data is pre-processed. The customer number, age, income level, years of work, family size, asset value, property value, health expenses, insurance amount, account balance, repayment record, purchase frequency, deposit amount, credit card overdraft limit, vehicle value, and loan interest rate are used as feature variables. The feature variables are obtained through data collection forms, financial institutions, and online questionnaires. The credit score is used as the target variable and is obtained through financial institutions.

[0070] The obtained data are processed for outliers, missing values, duplicates and data standardization. When a group of customer data has too many missing values, this group of data can be directly deleted to avoid affecting the effect of the overall customer data. When a group of customer data has fewer missing values, the mean value filling method can be selected for filling. The box plot method is used to detect outliers. The first quartile is (Q1), the second quartile is (Q2), and the third quartile is (Q3), then IQR = (Q3 - Q1). Outliers are defined as those data points that are less than (Q1 - 1.5*IQR) or greater than (Q3 + 1.5*IQR). The median value is used to replace the outliers to avoid deleting outliers.

[0071] Perform consistency checks on the preprocessed data, including data type consistency checks and data range consistency checks. Data type consistency checks ensure that the data type of each variable is correct; range consistency is used to check whether the value of each numeric variable is within a reasonable range to avoid the introduction of new abnormal values ​​or values ​​outside the normal range during data processing.

[0072] Too many features will lead to data redundancy and poor performance in practical applications, while irrelevant or redundant features will introduce noise. Feature selection is performed on the data collected by the customer collection system. The feature vector is , where each For one of the features, the importance between the feature variable (customer information) and the target variable (customer credit score) is analyzed using the random forest importance analysis method to obtain the feature importance vector , use the Lasso method to analyze the correlation between the feature variable (customer information) and the target variable (customer need score), and obtain the regression coefficient of each feature through the Lasso method ,calculate The absolute value of the coefficient =( ,...., )};

[0073] Calculate the importance score of random forest and the absolute value of regression coefficient of Lasso method for weighted average to get the comprehensive value of customer characteristics , the formula is as follows:

[0074]

[0075] Among them, p is 0.5, k is the index of the kth feature, and then calculate the 80th percentile threshold T of the feature comprehensive value, and select The features are taken as the final features, and the final features are determined after analysis. These features are considered to be the most valuable information for predicting customer credit scores.

[0076] After preprocessing and feature selection, the final features are used as input, and the SVR model is selected for modeling. The data set is randomly divided into a training set and a test set according to a ratio of 7:3 for training. The appropriate kernel function (linear, polynomial, RBF), penalty parameter C and gamma parameters are selected through grid search. In SVR, for a given training data set Where xi is the input feature, is the corresponding target value. The goal of the SVR model is to find a prediction function , the formula is as follows:

[0077]

[0078] Where w is the weight vector, T is the transposed sign, b is the bias coefficient, Yes Input The feature map of The result after mapping to high-dimensional space, kernel function Provides a measure of the similarity or distance between two samples, where represents the input feature vector of the i-th sample in the training data set, Represents the input feature vector of the jth sample in the training dataset.

[0079] Establishment of optimization problem: SVR optimizes the model by minimizing the loss function and regularization term. In order to determine w and b, SVR optimizes the following objective problem as shown below:

[0080]

[0081] in, is a regularization term used to prevent overfitting and ensure the simplicity of the model. C is a penalty factor that weighs the impact of regularization and loss function. and are slack variables for the upper and lower bounds, and the constraints are:

[0082]

[0083] in, is the insensitive loss factor.

[0084] Lagrange dual problem: The Lagrange multiplier is introduced to make it easier to solve. The dual problem is as follows:

[0085]

[0086] The constraints are:

[0087]

[0088] Among them, the introduction of two sample indexes i and j is used to construct the optimization problem. , is the Lagrange multiplier of sample i , is the Lagrange multiplier of sample j, and the kernel function is used to describe and optimize the relationship between samples in the training data set.

[0089] Solve the dual problem: By optimizing the dual problem above, we can solve the Lagrange multipliers, and then calculate w and b based on these multipliers. The formula is as follows:

[0090]

[0091]

[0092] Through the above process, w and b are calculated, and the test set data is used to evaluate the performance of the model. The evaluation indicators used are determination coefficient, mean square error, root mean square error, and mean absolute error. The determination coefficient reflects the proportion of the variance of the predicted data explained by the model. The closer the value is to 1, the stronger the explanatory ability is. The mean square error and root mean square error represent the average of the square difference between the predicted value and the true value and its square root, respectively. The smaller the two are, the higher the prediction accuracy is. The mean absolute error measures the average absolute value of the prediction error. The smaller the value is, the higher the prediction accuracy of the model is. These indicators evaluate the accuracy, explanatory power and error distribution of the model. The specific formulas are as follows:

[0093]

[0094]

[0095]

[0096]

[0097] in, is the actual value, is the predicted value, is the average of the actual values. The comprehensive evaluation index D is used to understand the customer credit score. The formula is as follows:

[0098]

[0099] Among them, when w is 0.25, the comprehensive evaluation index 1≤ When <10, the fitting effect is better and the model predicts the credit score more accurately; 10≤ When D <50, the fitting effect is average and the predicted credit score result is average; when D ≥50, the model predicts the credit score result poorly. When the SVR model predicts the customer's credit score, if the comprehensive evaluation index is average or poor, it means that the model prediction is inaccurate and the model's hyperparameters need to be adjusted to improve the model's accuracy.

[0100] When the customer data analysis and processing module is completed, the final required results are obtained, and data visualization intuitively understands and presents the data results. Specifically, it provides the following functions:

[0101] Generate charts: The data visualization module generates various types of charts, including line charts, bar charts, scatter charts, and pie charts, which intuitively display the distribution, trends, and relationships of data;

[0102] Visual reports: The data visualization module generates interactive visual reports, including tables and text descriptions. You can browse the reports to obtain data analysis results and interact with the data.

[0103] Data dashboard: The data visualization module creates a data dashboard to display real-time changes in key indicators and business indicators. The dashboard allows users to quickly understand the status and trends of data and make timely decisions.

[0104] See also Figure 2 The present invention provides a method for backend management of customer data, comprising the following steps:

[0105] Used for customer identity authentication and authorization, where multi-factor authentication includes password verification and SMS verification code to ensure the security and accuracy of customer identity authentication;

[0106] Used to obtain customer information, including collecting customer number, age, income level, years of work experience and number of family members through data collection forms; obtaining credit scores, asset values, property values, health expenses and insurance amounts through financial institutions; obtaining customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values ​​and loan interest rates through online questionnaires; updating customer data through real-time event-driven mechanisms and regular polling to ensure the comprehensiveness and timeliness of key customer information;

[0107] It is used to receive data from the data collection module. When the customer data is updated, it uses the Ethereum blockchain platform and smart contracts to manage data updates. It then uses the PoS (Proof of Stake) consensus mechanism to determine the consistency of the updated data, encrypt the data, and save it in the MySQL database. It implements regular data backup strategies to improve data security and durability.

[0108] Used to analyze and process stored customer data information, including data preprocessing, feature selection, and support vector regression (SVR) modeling, predict customer credit scores, and make more accurate risk assessments and decisions;

[0109] Used to provide intuitive data results, generate visual reports, charts and data dashboards, and quickly understand data trends, discover patterns, and perform decision analysis to optimize customer service and business operations.

[0110] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0111] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the 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.

Claims

1. A customer data background management system, characterized by: Customer authentication module, used for customer identity authentication and authorization, where multi-factor authentication includes password verification and SMS verification code; The customer data collection module is used to obtain customer information. It updates customer data through real-time event-driven mechanisms and regular polling. It collects customer numbers, ages, income levels, years of work experience, and family size through data collection forms; obtains credit scores, asset values, property values, health expenses, and insurance amounts through financial institutions; and obtains customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values, and loan interest rates through online questionnaires. The customer data storage module is used to receive data from the data collection module. When the customer data is updated, it uses the Ethereum blockchain platform to manage data updates using smart contracts, and then uses the PoS consensus mechanism to determine the consistency of the updated data, encrypt the data, and save it in the MySQL database, implementing a regular data backup strategy; The customer data analysis and processing module is used to analyze and process the stored customer data information, including data preprocessing, feature selection and support vector regression (SVR) model building, and predict customer credit scores; Data visualization module, used to provide intuitive data results and generate visual reports, charts and data dashboards; The customer data collection module supports multiple data sources, including data collection forms, financial institutions and online questionnaires to obtain key customer information, including: customer number, age, income level, years of work, family size, credit score, asset value, property value, health expenses, insurance amount, account balance, repayment record, purchase frequency, deposit amount, credit card overdraft limit, vehicle value and loan interest rate. This module obtains key customer information from data collection forms, financial institutions and online questionnaires through real-time event-driven mechanisms and regular polling updates, and captures real-time updated data and updates it to the customer data storage module; The customer data analysis and processing module performs preprocessing and feature selection on the customer data stored in the customer data storage module. The data collected by the data acquisition module includes linear data and nonlinear data. Taking into account different data characteristics, the random forest importance analysis method and the Lasso method are used for feature selection to comprehensively analyze the correlation between customer data and credit scores, and further select a suitable feature set. The feature vector is , where each For one of the features, k ranges from 1 to 16; Use the random forest importance analysis method to analyze the importance between the feature variable and the target variable. The feature variable is customer information and the target variable is the customer's credit score. Get the feature importance vector , use the Lasso method to analyze the correlation between the feature variable customer information and the target variable customer need score, and obtain the regression coefficient of each feature through the Lasso method ,calculate The absolute value of the coefficient =( ,...., )}; Calculate the importance score of random forest and the absolute value of regression coefficient of Lasso method for weighted average to get the comprehensive value of customer characteristics , the formula is as follows: Among them, p is 0.5, k is the index of the kth feature, and then calculate the 80th percentile threshold T of the feature comprehensive value, and select The features of .

2. A customer data background management system according to claim 1, characterized in that: In SVR, for a given training dataset in are input features, is the corresponding output feature. The goal of the SVR model is to find a prediction function , the formula is as follows: Where w is the weight vector, T is the transposed sign, b is the bias coefficient, Yes Input The feature map of The result after mapping to high-dimensional space, kernel function Provides a measure of the similarity or distance between two samples, where represents the input feature vector of the i-th sample in the training data set, Represents the input feature vector of the jth sample in the training data set, and N is the number of training data samples.

3. A customer data background management system according to claim 2, characterized in that: SVR optimizes the model by minimizing the loss function and regularization term, determining w and b. SVR optimizes the following objective problem as shown below: in, is a regularization term used to prevent overfitting and ensure the simplicity of the model. C is a penalty factor that weighs the impact of regularization and loss function. and are slack variables for the upper and lower bounds, and the constraints are: in, In order to be insensitive to the loss factor, the Lagrange multiplier is introduced, which makes it easier to solve. The dual problem is as follows: The constraints are: Among them, the introduction of two sample indexes i and j is used to construct the optimization problem. , is the Lagrange multiplier of sample i , is the Lagrange multiplier of sample j, and the kernel function is used to describe and optimize the relationship between samples in the training data set.

4. A customer data background management system according to claim 3, characterized in that: The Lagrange multiplier is obtained by solving the above optimization dual problem, and then w and b are calculated based on the multiplier. The formula is as follows: According to the above formula, w and b can be obtained.

5. A customer data background management system according to claim 4, characterized in that: The evaluation indicators used are determination coefficient, mean square error, root mean square error, and mean absolute error. The determination coefficient reflects the proportion of the variance of the model to explain the predicted data. The closer the value is to 1, the stronger the explanatory power is. The mean square error and root mean square error represent the average of the square difference between the predicted value and the true value and its square root, respectively. The smaller the two are, the higher the prediction accuracy is. The mean absolute error measures the average absolute value of the prediction error. The smaller the value is, the higher the prediction accuracy of the model is. These indicators evaluate the accuracy, explanatory power, and error distribution of the model. The specific formulas are as follows: in, is the actual value, is the predicted value, is the average of the actual values. The comprehensive evaluation index D is used to understand the customer credit score. The formula is as follows: Among them, when w is 0.25, the comprehensive evaluation index 1≤ When <10, the fitting effect is better and the model predicts the credit score more accurately; 10≤ When D <50, the fitting effect is average and the predicted credit score result is average; when D ≥50, the model predicts the credit score result poorly. When the SVR model predicts the customer's credit score, if the evaluation index effect is average or poor, it means that the model prediction is inaccurate and the model's hyperparameters need to be adjusted to improve the model's accuracy.

6. A method for using customer data background management, applied to a customer data background management system in claim 1, characterized in that: Used for customer identity authentication and authorization, where multi-factor authentication includes password verification and SMS verification code; Used to obtain customer information, update customer data through real-time event-driven mechanisms and regular polling, collect customer numbers, ages, income levels, years of work experience, and family size through data collection forms; obtain credit scores, asset values, property values, health expenses, and insurance amounts through financial institutions; obtain customers' account balances, repayment records, purchase frequency, deposit amounts, credit card overdraft limits, vehicle values, and loan interest rates through online questionnaires; It is used to receive data from the data collection module. When the customer data is updated, it uses the Ethereum blockchain platform to manage data updates using smart contracts, and then uses the PoS consensus mechanism to determine the consistency of the updated data. It also encrypts the data and saves it in the MySQL database, implementing a regular data backup strategy. Used to analyze and process stored customer data information, including data preprocessing, feature selection, and support vector regression (SVR) model building to predict customer credit scores; Used to provide intuitive data results and generate visual reports, charts and data dashboards.

Images