Potential customer mining method and system based on data mining

Abstract

The invention provides a potential customer mining method and system based on data mining, and relates to the technical field of data mining, and the method comprises the steps: 1, collecting customer related data, including basic information, behavior data, transaction data and customer feedback data of customers; 2, cleaning, converting and integrating the collected data; according to the invention, through deep analysis of multi-dimensional customer data, customer groups with high potential purchase possibility can be accurately identified, blindness of a traditional marketing mode is avoided, an enterprise can intensively invest marketing resources to customers who are most likely to generate rewards, and through a prediction result based on a data mining model, the marketing efficiency of the enterprise is improved. According to the invention, enterprises can formulate personalized marketing strategies, accurately push product information and customize preferential activities according to potential customers, and the response rate of the customers to the marketing activities is improved, so that the sales cycle is shortened, and the marketing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a potential customer mining method and system based on data mining. Background Art

[0002] In today's highly competitive business environment, companies are facing increasing market pressure and need to constantly find new customers to expand their business scale and increase their market share. Traditional customer acquisition methods, advertising, and marketing activities often have high costs and low efficiency, making it difficult to accurately locate customers with real potential needs. With the rapid development of information technology, companies have accumulated a large amount of customer-related data. These data contain rich information, including customer behavior, preferences, and needs. How to effectively use this data to tap potential customers has become the key to improving the competitiveness of companies. Data mining technology, as an effective means of discovering valuable information from massive data, provides a possible solution to this problem. Through data mining, customer data can be deeply analyzed to discover the patterns and laws hidden behind the data, thereby predicting which customers have a higher potential to purchase.

[0003] Compared with traditional methods, potential customer mining methods based on data mining can locate target customers more accurately, improve the utilization efficiency of marketing resources, reduce marketing costs, increase customer conversion rate and corporate profits. However, the existing potential customer mining methods and systems still have some shortcomings. Some methods do not make full use of data and only consider some customer characteristics, resulting in low accuracy of mining results. Some systems have poor scalability and adaptability and are difficult to cope with data of different types and sizes. Some methods are not flexible enough in model construction and algorithm selection, and cannot be optimized according to the specific business needs of the enterprise.

[0004] Therefore, it is of great practical significance to develop an efficient, accurate, flexible and well-scalable potential customer mining method and system based on data mining. Summary of the Invention

[0005] To achieve the above objectives, the present invention proposes a potential customer mining method and system based on data mining, comprising:

[0006] Step 1: Collect customer-related data, including basic customer information, behavioral data, transaction data, and customer feedback data;

[0007] Step 2: Clean, transform and integrate the collected data;

[0008] Step 3: Extract meaningful features from the processed data, including calculating customer behavioral features, constructing business rule-based features, and sentiment tendency features;

[0009] Step 4: Select an algorithm to build a potential customer prediction model, train the model using the training set, and adjust the model parameters to fit the potential customer patterns in the training data;

[0010] Step 5: Evaluate the model using the following metrics: accuracy, recall, F1 value, and area under the ROC curve.

[0011] Step 6: Input new customer data into the evaluated and optimized model to obtain the probability or score of each customer becoming a potential customer;

[0012] Step 7: Analyze the potential customers discovered, understand their characteristics and behavior patterns, and develop personalized marketing strategies.

[0013] In one example, the customer's basic information includes age, gender, occupation, contact information, purchase history, clickstream data, social media interactions, transaction data, order amount, purchase frequency, and purchase time. Customer feedback data includes evaluations, complaints, and suggestions. Data is cleaned to remove noise, duplicate data, and erroneous data. Classification data is encoded, numerical data is standardized, and data from different data sources is integrated to ensure data consistency and integrity.

[0014] In one example, the behavioral characteristics include the average purchase interval, the time of the most recent purchase, the average dwell time on the browsed page, and the diversity of purchased categories. A feature subset with high predictive power for potential customer mining is screened out, and the data processed by feature engineering is divided into a training set and a test set, and the training set is used to train the model.

[0015] In one example, the model evaluation and optimization uses a test set to evaluate the trained model, analyze the model's prediction results, identify incorrectly predicted samples, further adjust the model's parameters, add more features, try different algorithms, and optimize the model's performance. Customer prediction and screening sets a threshold based on business needs, and identifies customers with scores above the threshold as potential customers. These potential customers can be further sorted by score.

[0016] In one example, the result analysis and application include analyzing the mined potential customers, understanding their characteristics and behavior patterns, formulating personalized marketing strategies, designing different marketing activities for different types of potential customers, pushing personalized product recommendations, tracking the conversion of potential customers, and feeding back the actual converted customer data into the model to further optimize and improve the model.

[0017] A potential customer mining system based on data mining, characterized by comprising:

[0018] Data collection module: responsible for collecting customer-related data from multiple data sources;

[0019] Data preprocessing module: performs cleaning, conversion and integration preprocessing operations on the collected data;

[0020] Feature Engineering Module: This module implements the extraction, construction, and selection of features. It generates a representative and predictive feature set based on business needs and data characteristics, providing effective input features for model construction.

[0021] Model training and management module: provides model training, evaluation, optimization, and storage functions, and supports a variety of machine learning and data mining algorithms;

[0022] Potential customer prediction module: inputs new customer data into the trained model to predict and score potential customers;

[0023] The result display and analysis module provides data analysis functions to help users gain a deeper understanding of the characteristics and behaviors of potential customers and provide decision support for formulating marketing strategies;

[0024] The marketing decision support module recommends appropriate marketing channels, marketing content and marketing timing.

[0025] In one example, the relevant data includes database interfaces, web crawlers, and API interfaces to achieve automatic data collection and updating. The preprocessing operations ensure the quality and consistency of the data, providing a reliable data foundation for subsequent analysis and modeling. The feature engineering module generates a representative and predictive feature set based on business needs and data characteristics, providing effective input features for model construction.

[0026] In one example, the model training and management module can easily switch and adjust algorithms and parameters, manage different versions of models, and record the model training process and performance indicators.

[0027] In one example, the potential customer prediction module inputs new customer data into the trained model, predicts and scores potential customers, and outputs a list of potential customers and their related information, predicted probabilities, and customer characteristics.

[0028] The potential customer mining method and system based on data mining proposed by the present invention can bring the following beneficial effects:

[0029] 1. The present invention can accurately identify customer groups with high potential purchasing possibility through in-depth analysis of multi-dimensional customer data, avoiding the blindness of traditional marketing methods, and enabling enterprises to concentrate marketing resources on customers who are most likely to generate returns. Through the prediction results based on data mining models, enterprises can formulate personalized marketing strategies for potential customers, accurately push product information, customize promotional activities, and increase customer response rate to marketing activities, thereby shortening the sales cycle and improving marketing efficiency.

[0030] 2. The present invention can reduce unnecessary marketing expenses through accurate potential customer mining and avoid wasting resources on non-potential customer groups. Enterprises can allocate marketing budgets more reasonably and invest funds in more valuable marketing channels and activities, thereby reducing overall marketing costs. Through personalized marketing methods, it can better meet the personalized needs of customers, allowing customers to feel the company's attention and attention, thereby improving customer satisfaction and loyalty to the company and establishing a good brand image for the company.

[0031] 3. The detailed data analysis and potential customer insights provided by the method and system of the present invention can provide powerful decision-making support for the management of the enterprise. The enterprise can formulate more scientific product development strategies, marketing plans and sales targets according to the characteristics of potential customers and market trends, thereby improving the overall competitiveness of the enterprise. The system has good scalability and flexibility. With the development of the enterprise's business and the continuous enrichment of data, new data sources, feature variables and model algorithms can be easily added to adapt to business needs and market changes at different stages, and maintain the accuracy and effectiveness of potential customer mining. In the process of mining potential customers, through comprehensive analysis of data, the enterprise may discover some customer groups or market segments that have not been paid attention to before, providing clues for the enterprise to open up new market opportunities and promote the diversified development of the enterprise's business. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0033] Figure 1 The figure is a flow chart of a potential customer mining method based on data mining according to the present invention.

[0034] Figure 2 The figure is a structural diagram of a potential customer mining system based on data mining according to the present invention.

[0035] Figure 3 This is a schematic diagram of feature selection of a potential customer mining method based on data mining in the present invention. DETAILED DESCRIPTION

[0036] In order to more clearly and completely illustrate the technical solution of the present invention, the present invention will be further described below with reference to the accompanying drawings.

[0037] Please refer to Figure 1-Figure 3 The present invention proposes a specific implementation method of a potential customer mining method and system based on data mining, data collection and preprocessing, and collects multi-dimensional data of customers through the company's internal database, sales management system, customer relationship management system, external data platform, industry data service provider, and social media platform channels. The collected data types include but are not limited to customer basic information, age, gender, contact information, transaction data, purchase time, purchase amount, purchase frequency, behavioral data, website browsing time, click behavior, search keywords, as well as feedback data, product evaluation, and complaint records.

[0038] The collected data is imported into the data cleaning module, and the data cleaning algorithm is used to remove duplicate data, outliers and missing values. For missing values, mean filling and regression prediction methods are used for processing. For outliers, they are identified and corrected or deleted through tools such as box plots. After the cleaning is completed, the data is standardized and normalized to make different types of data comparable. Finally, the processed data is stored in the data warehouse.

[0039] Feature engineering and feature extraction extract various features from preprocessed data. For example, features such as the time interval between customers' recent purchases, average purchase amount, and number of purchased categories can be extracted from transaction data. Features such as the average daily website visit duration, number of page jumps, and search keyword popularity can be extracted from behavioral data. Natural language processing technology is used to segment and tag customer feedback data, extracting features such as high-frequency keywords and sentiment tendency scores.

[0040] Feature selection: Use feature selection algorithms such as chi-square test and mutual information to calculate the correlation between each feature and whether the customer is a potential customer, screen out features with high correlation, remove redundant and irrelevant features, and form the final feature set.

[0041] Model building and training, model selection: Based on data characteristics and business needs, select an appropriate machine learning model, such as random forest, support vector machine, or neural network. For example, if the data scale is large and the features are complex, a neural network model can be selected. If the data has nonlinear relationships, a support vector machine may be more appropriate.

[0042] Model training: Divide the feature set into a training set and a test set, and use the training set to train the selected model. During the training process, set the model's hyperparameters, the number of layers and nodes of the neural network, and the number of trees in the random forest. Adjust the hyperparameters through cross-validation to achieve optimal performance on the training set.

[0043] Model evaluation and optimization, Model evaluation: Use the test set to evaluate the trained model and calculate the accuracy, recall, F1 value, and AUC evaluation indicators. The accuracy reflects the proportion of potential customers correctly predicted by the model, and the recall measures the proportion of potential customers identified by the model to actual potential customers.

[0044] Model optimization: Analyze the shortcomings of the model based on the evaluation results. If the model has an overfitting problem, it can be optimized by regularization, reducing model complexity, etc. If it has an underfitting problem, it can be optimized by adding training data, adjusting model parameters, or replacing the model. Repeat the model training, evaluation, and optimization process until the model performance meets the expected requirements.

[0045] Potential customer prediction and screening: New customer data is input into the optimized model. The model outputs the probability score of each customer becoming a potential customer. A probability threshold is set, and customers with scores above the threshold are screened out as potential customers that the company focuses on. They are then sorted according to their scores to facilitate targeted marketing by the company.

[0046] Results Analysis and Application: Conduct in-depth analysis of the discovered potential customers and use visualization tools such as bar charts, line charts, and heat maps to display information such as the characteristic distribution and behavioral patterns of potential customers. For example, a bar chart can be used to display the number distribution of potential customers in different age groups, and a line chart can be used to analyze the changing trend of potential customers' purchase frequency over time. Based on the analysis results, companies can develop personalized marketing strategies, such as pushing information about new fashion products to young potential customers with high purchase frequency, and providing exclusive promotions to potential customers with high consumption potential.

[0047] Specific implementation method of the system The potential customer mining system based on data mining of the present invention includes a data acquisition module, a data preprocessing module, a feature engineering module, a model training and management module, a potential customer prediction module, a result display and analysis module and a marketing decision support module.

[0048] Data collection module: This module interacts with data sources through API interfaces, database connections, etc. to achieve automatic data collection and regular updates. It establishes an API connection with the enterprise CRM system to obtain the latest customer transaction and behavior data in real time, and regularly downloads industry data and market trend data from external data platforms.

[0049] Data preprocessing module The data preprocessing module receives the collected data and automatically performs data cleaning, conversion and integration operations. By writing data cleaning rules and scripts, it identifies and processes duplicate data, missing values and outliers. It uses data conversion tools to unify the data format and convert the date format into a standard format. Through data integration technology, it integrates data from different data sources and stores them in the data warehouse to provide a unified data view for subsequent analysis.

[0050] Feature Engineering Module The feature engineering module provides a visual feature extraction and selection operation interface. Users can select the feature type and algorithm to be extracted through the interface. The system automatically extracts the corresponding features from the data warehouse and performs feature selection operations. After the operation is completed, the final feature set is stored in the model training data set for subsequent model training.

[0051] The model training and management module integrates a variety of machine learning algorithm libraries. Users can select models and set hyperparameters through a graphical interface. During the model training process, the system monitors the training progress and performance indicators in real time and records the training log. After the training is completed, the model is saved to the model library. Users can perform operations such as version management, loading and evaluation of the model.

[0052] Potential Customer Prediction Module The potential customer prediction module is connected to the model library and data warehouse. When there is new customer data, it automatically obtains data from the data warehouse, calls the optimal model in the model library for prediction, outputs a list of potential customers and their probability scores, and stores the results in the result database.

[0053] The result display and analysis module reads potential customer data from the result database and uses visualization tools to generate various charts, such as potential customer distribution maps and feature comparison charts. At the same time, it provides data analysis functions, allowing users to deeply analyze the characteristics and behavior patterns of potential customers through operations such as filtering, sorting, and drilling.

[0054] Marketing Decision Support Module The Marketing Decision Support Module generates personalized marketing strategy recommendations based on the analysis results of potential customers, combined with the company's marketing goals and resources, through preset decision rules and algorithms. For example, based on the potential customers' purchasing preferences and consumption capacity, it recommends appropriate product combinations and marketing channels, and determines the optimal time and location for marketing activities based on the potential customers' geographical location and time characteristics.

[0055] In this embodiment, the present invention achieves accurate mining of potential customers through a specific, operational process and system structure. Data collection and preprocessing: Collect multi-dimensional customer data from multiple data sources, including basic information, transaction data, behavioral data, and feedback data. The collected data is imported into a data cleaning module. Data cleaning algorithms are used to remove duplicate data, outliers, and missing values. Missing values are filled or predicted, and outliers are corrected or deleted. The cleaned data is standardized and normalized to make different types of data comparable. The processed data is then stored in a data warehouse. Feature engineering: Extract various features from the preprocessed data, such as purchase time interval, average purchase amount, and number of purchased categories in transaction data; and average daily website visit duration, page jump number, and search keyword popularity in behavioral data. Natural language processing technology is used to segment and tag customer feedback data, extracting features such as high-frequency keywords and sentiment tendency scores. Feature selection algorithms such as chi-square test and mutual information are used to screen out features that are highly correlated with whether the customer is a potential customer, thereby forming a final feature set.

[0056] Model Building and Training: Select an appropriate machine learning model, such as a random forest, support vector machine, or neural network, based on data characteristics and business needs. Divide the feature set into a training set and a test set. Use the training set to train the selected model. During training, set the model's hyperparameters and adjust them through cross-validation. Model Evaluation and Optimization: Use the test set to evaluate the trained model, calculating metrics such as accuracy, recall, F1 score, and AUC. Based on the evaluation results, analyze the model's shortcomings and take appropriate optimization measures, such as regularization, adding training data, adjusting model parameters, or replacing the model. Lead Customer Prediction and Screening: Input new customer data into the optimized model. The model outputs a probability score for each customer becoming a potential customer. A probability threshold is set, and customers with scores above the threshold are selected as potential customers for the company's key focus. They are then ranked by score. Result Analysis and Application: Conduct in-depth analysis of the discovered potential customers, using visualization tools to display information such as their characteristic distribution and behavior patterns. Based on the analysis results, the company can develop personalized marketing strategies, such as delivering customized product recommendations or promotions to potential customers with specific characteristics.

[0057] Specific implementation methods of the system: Data acquisition module: interacts with the data source through API interface, database connection, etc. to realize automatic data collection and regular update; Data preprocessing module: receives the collected data, performs data cleaning, conversion and integration operations, and provides a unified data view for subsequent analysis; Feature engineering module: provides a visual feature extraction and selection operation interface, and users can select the feature type and algorithm to be extracted through the interface; Model training and management module: integrates a variety of machine learning algorithm libraries, and users can select models and set hyperparameters through a graphical interface; Potential customer prediction module: connected to the model library and data warehouse, automatically makes predictions when there is new customer data, and outputs a list of potential customers and their probability scores; Result display and analysis module: reads potential customer data from the result database, uses visualization tools to generate various charts, and provides data analysis functions; Marketing decision support module: generates personalized marketing strategy recommendations based on the analysis results of potential customers to help companies make more scientific marketing decisions.

[0058] Of course, the present invention may have many other implementations. Based on this implementation, other implementations obtained by ordinary technicians in this field without any creative work are all within the scope of protection of the present invention.

Claims

1. A potential customer mining method based on data mining, characterized in that: include: Step 1: Collect customer-related data, including basic customer information, behavioral data, transaction data, and customer feedback data; Step 2: Clean, transform and integrate the collected data; Step 3: Extract meaningful features from the processed data, including calculating customer behavioral features, constructing business rule-based features, and sentiment tendency features; Step 4: Select an algorithm to build a potential customer prediction model, train the model using the training set, and adjust the model parameters to fit the potential customer patterns in the training data; Step 5: Evaluate the model using the following metrics: accuracy, recall, F1 value, and area under the ROC curve. Step 6: Input new customer data into the evaluated and optimized model to obtain the probability or score of each customer becoming a potential customer; Step 7: Analyze the potential customers discovered, understand their characteristics and behavior patterns, and develop personalized marketing strategies.

2. A method and system for mining potential customers based on data mining according to claim 1, characterized in that: The customer's basic information includes age, gender, occupation, contact information, purchase history, clickstream data, social media interactions, transaction data, order amount, purchase frequency, and purchase time. Customer feedback data includes evaluations, complaints, and suggestions. Data is cleaned to remove noise, duplicate data, and erroneous data. Categorical data is encoded, numerical data is standardized, and data from different data sources is integrated to ensure data consistency and integrity.

3. The method and system for mining potential customers based on data mining according to claim 1, characterized in that: The behavioral characteristics include the average purchase interval, the time of the most recent purchase, the average dwell time on the browsed page, and the diversity of purchased categories. Feature subsets with high predictive power for potential customer mining are screened out, and the data processed by feature engineering is divided into training sets and test sets. The training set is used to train the model.

4. A method and system for mining potential customers based on data mining according to claim 1, characterized in that: The model evaluation and optimization uses a test set to evaluate the trained model, analyze the model's prediction results, identify incorrectly predicted samples, further adjust the model's parameters, add more features, try different algorithms, and optimize the model's performance. Customer prediction and screening sets a threshold based on business needs, and identifies customers with scores above the threshold as potential customers. These potential customers can be further sorted by score.

5. The method and system for mining potential customers based on data mining according to claim 1, characterized in that: The result analysis and application include analyzing the potential customers mined, understanding their characteristics and behavior patterns, formulating personalized marketing strategies, designing different marketing activities for different types of potential customers, pushing personalized product recommendations, tracking the conversion of potential customers, feeding back the actual converted customer data into the model, and further optimizing and improving the model.

6. A potential customer mining system based on data mining, characterized in that: include: Data collection module: responsible for collecting customer-related data from multiple data sources; Data preprocessing module: performs cleaning, conversion and integration preprocessing operations on the collected data; Feature Engineering Module: Implement feature extraction, construction, and selection functions, generate representative and predictive feature sets based on business needs and data characteristics, and provide effective input features for model construction; Model training and management module: provides model training, evaluation, optimization, and storage functions, and supports a variety of machine learning and data mining algorithms; Potential customer prediction module: inputs new customer data into the trained model to predict and score potential customers; The result display and analysis module provides data analysis functions to help users gain a deeper understanding of the characteristics and behaviors of potential customers and provide decision support for formulating marketing strategies; The marketing decision support module recommends appropriate marketing channels, marketing content and marketing timing.

7. A potential customer mining system based on data mining according to claim 6, characterized in that: The relevant data includes database interfaces, web crawlers, and API interfaces to achieve automatic data collection and updating. The preprocessing operations ensure the quality and consistency of the data, providing a reliable data foundation for subsequent analysis and modeling. The feature engineering module generates a representative and predictive feature set based on business needs and data characteristics, providing effective input features for model construction.

8. The potential customer mining system based on data mining according to claim 6, characterized in that: The model training and management module can easily switch and adjust algorithms and parameters, manage different versions of models, and record the model training process and performance indicators.

9. The potential customer mining system based on data mining according to claim 6, characterized in that: The potential customer prediction module inputs new customer data into the trained model, predicts and scores potential customers, and outputs a list of potential customers and their related information, predicted probability, and customer characteristics.

Images