Advertisement marketing system for after-sales data statistics

Abstract

The invention relates to the technical field of advertisement marketing, in particular to an advertisement marketing system for after-sales data statistics. The system comprises a data acquisition module, an intelligent analysis module, a precision marketing module and an evaluation feedback module. The customer basic information, purchase information, after-sales behavior data and customer feedback data are acquired through the data acquisition module, the intelligent analysis module establishes a recurrent neural network model for predicting the customer loss probability, and the precision marketing module performs precision marketing by taking the customer loss probability predicted by the intelligent analysis module as the horizontal axis and the customer value as the longitudinal axis. The method comprises the following steps: establishing a probability-divided two-dimensional matrix, formulating a corresponding advertisement marketing strategy, adopting different advertisement marketing strategies for different types of customers, improving pertinence, redeeming the customers to the greatest extent, reducing enterprise loss, establishing an evaluation system by an evaluation feedback module, evaluating the effect of advertisement marketing activities at regular intervals, and improving the efficiency of advertisement marketing. And the advertisement marketing strategy is continuously optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of advertising marketing, and more specifically, to an advertising marketing system for after-sales data statistics. Background Art

[0002] Advertising marketing refers to the activity that an enterprise promotes its products through advertisements, facilitates direct purchases by consumers, expands product sales, and enhances the enterprise's popularity, reputation, and influence. With the rapid development of economic globalization and the market economy, advertising marketing activities play an increasingly important role in an enterprise's marketing strategy and are an important part of an enterprise's marketing mix.

[0003] Currently, enterprises expand the popularity of their products through advertising marketing strategies to attract more new customers, but lack advertising marketing for old customers, resulting in customer loss. In order to analyze the after-sales data statistics of old customers, establish a prediction model for customer loss probability, and formulate corresponding advertising marketing strategies for customers with different loss probabilities, combined with the potential value of customers, to maximize the recovery of customers and reduce the losses of enterprises. Therefore, we propose an advertising marketing system for after-sales data statistics. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem that enterprises lack advertising marketing for old customers, resulting in customer loss. In order to analyze the after-sales data statistics of old customers, establish a prediction model for customer loss probability, and formulate corresponding advertising marketing strategies for customers with different loss probabilities, combined with the potential value of customers, to maximize the recovery of customers and reduce the losses of enterprises.

[0005] To achieve the above purpose, the present invention provides an advertising marketing system for after-sales data statistics, including a data collection module, an intelligent analysis module, a precision marketing module, and an evaluation and feedback module;

[0006] The data collection module collects customer basic information, purchase information, after-sales behavior data, and customer feedback data from the enterprise's after-sales system and customer service records, cleans and labels the collected data, and transfers the processed data to the intelligent analysis module;

[0007] The intelligent analysis module uses the correlation analysis method to extract features from the data collected by the data collection module, and performs encoding and normalization processing on the extracted features. Taking the number of complaints and the interval between the last purchase time and the current time as input features and the predicted customer loss probability as the output feature, a recurrent neural network model for predicting the customer loss probability is established;

[0008] The precise marketing module uses the probability of customer churn predicted by the intelligent analysis module as the horizontal axis and customer value as the vertical axis to establish a two-dimensional matrix divided by probability, classifying customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers, and low-value high-churn customers. It uses the K-Means algorithm to cluster the customers within the quadrants and formulates corresponding advertising marketing strategies.

[0009] The evaluation and feedback module uses the customer response rate, purchase conversion rate, and customer retention rate as evaluation indicators to regularly evaluate the effectiveness of advertising marketing activities and dynamically adjust the advertising marketing strategies formulated by the precise marketing module according to the evaluation results.

[0010] As a further improvement of this technical solution, the intelligent analysis module includes a feature extraction unit and a model building unit.

[0011] The feature extraction unit uses the correlation analysis method to extract features from the collected data and arranges the extracted feature data in a time series.

[0012] The model building unit uses the ReLU function as the activation function for the hidden layer and the Softmax function as the activation function for the output layer to establish a neural network model for predicting the probability of customer churn, and uses the Adam optimizer to train the model.

[0013] As a further improvement of this technical solution, the feature extraction unit converts categorical features into numerical features through the label encoding method and maps the feature values to the range of 0 to 1 using the Z-score normalization method.

[0014] As a further improvement of this technical solution, the model building unit selects the multi-class cross-entropy loss function as the loss function of the model to measure the difference between the model prediction result and the real situation. Its formula is:

[0015] ;

[0016] where is the value of the loss function, is the number of classes, is the true class label, is the probability predicted to belong to the th class.

[0017] As a further improvement of this technical solution, the model building unit uses the grid search method to optimize the hyperparameters of the recurrent neural network model for predicting the customer churn probability, combines the values of each hyperparameter to form a grid-like parameter space, trains and evaluates the model for each set of parameter combinations in the parameter space, and selects the parameter combination with the best performance on the validation set as the optimal hyperparameters.

[0018] As a further improvement of this technical solution, the precision marketing module includes a probability division unit and a strategy formulation unit;

[0019] The probability division unit uses matrix analysis to divide customers according to the customer churn probability and customer value, and divides customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers, and low-value high-churn customers;

[0020] The strategy formulation unit uses the K-Means algorithm to cluster the customers within the quadrant, takes the customer's consumption amount and consumption frequency as the clustering features, and formulates corresponding advertising marketing strategies for the customers within the same cluster.

[0021] As a further improvement of this technical solution, the probability division unit uses the RFM model to establish a customer value evaluation system, and determines the customer value by calculating the customer's last consumption, consumption frequency, and consumption amount.

[0022] As a further improvement of this technical solution, the strategy formulation unit calculates the within-cluster sum of squared errors for different K values through the elbow method, plots the relationship curve between the K value and the within-cluster sum of squared errors, and determines the number of clusters.

[0023] As a further improvement of this technical solution, the evaluation and feedback module conducts a comparative analysis, compares the data of this advertising marketing activity with that of previous activities or the industry average level, finds out the advantages and gaps, and dynamically adjusts the advertising marketing strategy.

[0024] As a further improvement of this technical solution, the data acquisition module checks for missing values, duplicate values, and outliers in the data, and processes the data using the mean filling method.

[0025] Compared with the prior art, the beneficial effects of the present invention:

[0026] 1. In the advertising and marketing system for after-sales data statistics, the basic customer information, purchase information, after-sales behavior data, and customer feedback data are collected through the data collection module, and the data is processed. The intelligent analysis module extracts features from the processed data, uses the number of complaints and the interval between the last purchase time and the current time as input features, and uses the predicted customer churn probability as the output feature to establish a recurrent neural network model for predicting the customer churn probability, so as to realize predicting the customer churn probability based on the after-sales data of old customers;

[0027] 2. The precision marketing module takes the customer churn probability predicted by the intelligent analysis module as the horizontal axis and the customer value as the vertical axis to establish a two-dimensional matrix for probability division, divides customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers, and low-value high-churn customers, uses the K-Means algorithm to cluster the customers within the quadrant, and formulates corresponding advertising marketing strategies, adopts different advertising marketing strategies for different types of customers, improves the pertinence, maximally retrieves customers, and reduces the losses of the enterprise;

[0028] 3. The evaluation and feedback module takes the customer response rate, purchase conversion rate, and customer retention rate as evaluation indicators, regularly evaluates the effect of the advertising and marketing activities, dynamically adjusts the advertising marketing strategies formulated by the precision marketing module according to the evaluation results, continuously optimizes the advertising marketing strategies, and improves the resource utilization rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is the overall process schematic diagram of the present invention;

[0030] Figure 2 is the overall detailed process schematic diagram of the present invention.

[0031] The meanings of the various reference numerals in the figure are as follows:

[0032] 100, data collection module; 200, intelligent analysis module; 210, feature extraction unit; 220, model establishment unit; 300, precision marketing module; 310, probability division unit; 320, strategy formulation unit; 400, evaluation and feedback module. DETAILED DESCRIPTION OF THE INVENTION

[0033] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0034] At present, enterprises lack advertising and marketing for old customers, resulting in customer loss. In order to analyze the after-sales data statistics of old customers, establish a prediction model for customer loss probability, and formulate corresponding advertising and marketing strategies for customers with different loss probabilities, combined with the potential value of customers, so as to recover customers to the greatest extent and reduce the losses of enterprises.

[0035] Therefore, the present invention proposes to collect customer basic information, purchase information, after-sales behavior data and customer feedback data through a data collection module, process the data, establish a recurrent neural network model for predicting the customer loss probability by an intelligent analysis module, and use the customer loss probability predicted by the intelligent analysis module as the horizontal axis and the customer value as the vertical axis by a precision marketing module to establish a two-dimensional matrix for probability division, formulate corresponding advertising and marketing strategies, and evaluate and feedback module uses the customer response rate, purchase conversion rate and customer retention rate as evaluation indicators to regularly evaluate the effect of advertising and marketing activities and continuously optimize the advertising and marketing strategies.

[0036] Specifically as follows:

[0037] Please refer to Figure 1 As shown, the present invention provides an advertising and marketing system for after-sales data statistics, including a data collection module 100, an intelligent analysis module 200, a precision marketing module 300 and an evaluation and feedback module 400;

[0038] The data collection module 100 collects customer basic information, purchase information, after-sales behavior data and customer feedback data from the enterprise's after-sales system and customer service records, cleans and labels the collected data, and transfers the processed data to the intelligent analysis module 200;

[0039] The intelligent analysis module 200 uses the correlation analysis method to extract features from the data collected by the data collection module 100, encodes and normalizes the extracted features, uses the number of complaints and the interval between the last purchase time and the current time as input features, and the predicted customer loss probability as the output feature to establish a recurrent neural network model for predicting the customer loss probability;

[0040] The precision marketing module 300 uses the customer loss probability predicted by the intelligent analysis module 200 as the horizontal axis and the customer value as the vertical axis to establish a two-dimensional matrix for probability division, divides customers into high-value low-loss customers, high-value high-loss customers, low-value low-loss customers and low-value high-loss customers, uses the K-Means algorithm to cluster the customers within the quadrant, and formulates corresponding advertising and marketing strategies;

[0041] The evaluation feedback module 400 takes the customer response rate, purchase conversion rate, and customer retention rate as evaluation indicators, regularly evaluates the effectiveness of advertising and marketing activities, and dynamically adjusts the advertising marketing strategies formulated by the precision marketing module 300 according to the evaluation results.

[0042] Collect data related to customer after-sales from channels such as the enterprise's after-sales system and customer service records, including customer basic information (such as age, gender, region), purchase information (purchase time, purchase frequency, purchase amount), after-sales behavior data (number of consultations, number of complaints, repair records, return and exchange records), and customer feedback data (satisfaction score, evaluation content).

[0043] As Figure 2 shown, among them, the intelligent analysis module 200 includes a feature extraction unit 210 and a model establishment unit 220;

[0044] The feature extraction unit 210 uses the correlation analysis method to extract features from the collected data, and arranges the extracted feature data in time series;

[0045] The model establishment unit 220 uses the ReLU function as the activation function of the hidden layer, uses the Softmax function as the activation function of the output layer, establishes a neural network model for predicting the probability of customer churn, and uses the Adam optimizer to train the model;

[0046] The processed data is divided into a training set, a validation set, and a test set, divided according to a ratio of 7:2:1. The training set is used to train the model, the validation set is used to adjust the hyperparameters of the model and monitor the training process of the model to prevent overfitting, and the test set is used to evaluate the final performance of the model;

[0047] Input the training data into the neural network model, calculate the prediction result through forward propagation, then calculate the error between the prediction result and the true label according to the multi-class cross-entropy loss function, and then update the weights of the network through the backpropagation algorithm, continuously iterating this process until the loss function converges or reaches the preset number of training epochs.

[0048] In order to better extract features from the data, among them, the feature extraction unit 210 converts categorical features into numerical features through the label encoding method, and uses the Z-score normalization method to map the feature values to the range of 0 to 1;

[0049] Label encoding is a common method for converting categorical features into numerical features, which assigns a unique integer to each different category. Z-score normalization is a method for normalizing feature values, which converts the feature values into a distribution with a mean of 0 and a standard deviation of 1, and maps it to the range of 0 to 1;

[0050] First, it is necessary to calculate the mean and standard deviation of the feature vector. Using the calculated mean and standard deviation, perform Z-score normalization on each element in the original feature vector to obtain the normalized feature vector. For the normalized feature vector, find its minimum and maximum values, and use the method of linear transformation to map each element in the normalized feature vector to the range of 0 to 1, obtaining the finally mapped feature vector.

[0051] To better determine the loss function of the recurrent neural network model for predicting the churn probability of customers, among which, when establishing the model unit 220, the multi-class cross-entropy loss function is selected as the loss function of the model to measure the difference between the model prediction result and the actual situation, and its formula is:

[0052] ;

[0053] Among them, is the loss function value, is the number of classes, is the true class label, is the probability predicted to belong to the class.

[0054] Input the training set data into the constructed neural network model, perform forward propagation calculation according to the set network structure and hyperparameters, and through the linear transformation and activation function processing of each layer, obtain the predicted probability of the model for the customer churn category. According to the predicted probability output by the model and the true labels of the training set, use the multi-class cross-entropy loss function to calculate the loss value to measure the difference degree between the current prediction result of the model and the actual situation;

[0055] Through the backpropagation algorithm, backpropagate the loss value from the output layer to the input layer, calculate the gradient of each parameter to determine the direction and amplitude of parameter update. According to the calculated gradient, use the selected optimizer to update the parameters of the model according to a certain learning rate, so that the model is optimized in the direction of reducing the loss function. Repeat the above steps to perform multiple iterative trainings on the training set until the model converges or reaches the preset number of training epochs. During the training process, the performance of the model can be evaluated on the validation set regularly, observe the changes in indicators such as loss value, accuracy, and recall rate, and adjust the hyperparameters according to the results of the validation set, such as adjusting the learning rate, increasing or decreasing the number of hidden layer neurons, to prevent overfitting and improve the generalization ability of the model.

[0056] In order to better optimize the hyperparameters of the model, the model unit 220 is established using the grid search method to optimize the hyperparameters of the recurrent neural network model for predicting the customer churn probability. The values of each hyperparameter are combined to form a grid-like parameter space. The model is trained and evaluated for each set of parameter combinations in the parameter space, and the parameter combination that performs best on the validation set is selected as the optimal hyperparameter;

[0057] Suppose we want to optimize two hyperparameters, the learning rate and the number of neurons in the hidden layer, of a neural network. The value range of the learning rate is [0.001, 0.01, 0.1], and the value range of the number of neurons in the hidden layer is [32, 64, 128]. Then there will be 3×3 = 9 different parameter combinations. Train and evaluate these 9 combinations respectively, and select the optimal combination, which is the optimal hyperparameter.

[0058] Among them, the precision marketing module 300 includes a probability division unit 310 and a strategy formulation unit 320;

[0059] The probability division unit 310 uses matrix analysis to divide customers according to the customer churn probability and customer value, and divides customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers, and low-value high-churn customers;

[0060] The strategy formulation unit 320 uses the K-Means algorithm to cluster the customers within the quadrant, with the customer's consumption amount and consumption frequency as the clustering features, and formulates corresponding advertising marketing strategies for the customers within the same cluster;

[0061] Customers in different quadrants divided according to the two-dimensional matrix of customer churn probability and customer value have different characteristics and needs. Enterprises need to formulate targeted marketing strategies to maximize customer value and minimize churn risk. The following are specific strategies:

[0062] High-value low-churn probability customers: Provide exclusive personalized services for such customers, such as an exclusive customer service team, customized product or service solutions. Taking high-end credit card customers as an example, provide them with private banking services and exclusive airport VIP lounge services;

[0063] Launch higher-level membership benefits, such as doubled points, priority participation in new product trials or limited-edition product purchases, free access to high-end value-added services, etc., to further improve customer loyalty and satisfaction. Based on the customer's consumption preferences and behavior data, push highly accurate personalized marketing content and invite them to participate in high-end exclusive activities, such as high-end tasting events and private club activities held by the brand, to enhance the emotional connection between the customer and the brand;

[0064] High-value customers with a high probability of churn: Establish a dedicated customer retention team to conduct in-depth one-on-one communication with customers, understand the possible reasons for customer churn, and promptly address customer problems and pain points. For example, if a customer is wavering due to a competitor's low-price strategy, special price discounts or packages can be offered to them;

[0065] Based on the specific situation of the customer, develop personalized retention plans, such as providing exclusive discount offers, increasing service content, extending service periods, etc. When considering churn among high-value game players, limited game items, exclusive game packages, etc. can be provided to increase the frequency and depth of interaction with customers, conduct regular follow-ups, provide high-quality after-sales service, and hold exclusive customer appreciation activities to enhance customers' sense of identity and belonging to the brand;

[0066] Low-value customers with a low probability of churn: Adopt low-cost marketing channels, such as social media and SMS marketing, to regularly push product information, promotional activities, etc., maintain the brand's exposure in front of customers, and guide customers to try higher-value products or services by providing some low-threshold upgrade discounts or incentives, gradually increasing customer value. For example, an e-commerce platform provides full reduction coupons for ordinary users to encourage them to buy more or more expensive goods;

[0067] Incorporate such customers into the brand community, increase customer participation and stickiness through community activities, topic discussions, etc., cultivate customers' brand favorability, and create opportunities for their value improvement;

[0068] Low-value customers with a high probability of churn: Send targeted activation emails or SMS, provide special activation discounts, such as first-time consumption discounts, free gifts, etc., try to awaken dormant customers, conduct further value assessment of such customers, screen out customers with potential for value improvement, concentrate resources for key cultivation, and abandon those with extremely low value and high churn risk that are difficult to reduce;

[0069] Encourage customers to provide feedback, understand their dissatisfaction and needs regarding products or services, improve products or services based on the feedback, enhance customer satisfaction, and reduce the churn risk. If a customer feedbacks that a certain product has a cumbersome function, the product can be optimized and simplified.

[0070] In order to better determine customer value, among them, the probability division unit 310 uses the RFM model to establish a customer value assessment system, and determines customer value by calculating the customer's recency, frequency, and monetary value;

[0071] Recency: Refers to the number of days from the time when a customer last purchased a product or service to the present. This indicator reflects the customer's activity level. The closer the distance, the more likely the customer is to make a repeat purchase, and the higher the possible value to the enterprise;

[0072] Consumption Frequency: It refers to the number of times a customer purchases a product or service within a certain period. Customers with a high consumption frequency usually have a high degree of loyalty and recognition for the enterprise's products or services, and are important valuable customers of the enterprise;

[0073] Consumption Amount (Monetary): It refers to the total amount of money a customer spends on purchasing products or services within a certain period. This is an indicator that directly reflects the economic value created by the customer for the enterprise. The higher the consumption amount, the greater the customer value;

[0074] Separate scores are given to the three RFM indicators. Usually, customers are divided into different levels according to the distribution of the indicators. For example, each indicator is divided into 5 levels, with 1 point representing the lowest level and 5 points representing the highest level. Then, according to the RFM scores of each customer, they are divided into different value groups, such as important valuable customers (high R, high F, high M), important retained customers (low R, high F, high M), etc.;

[0075] According to the customer value assessment results, different marketing strategies are adopted for customers of different values. For important valuable customers, personalized high-quality services and exclusive offers are provided to maintain their loyalty. For potential value customers, targeted marketing plans are formulated to stimulate their consumption and enhance their value.

[0076] In order to better determine the number of clusters, among them, the strategy formulation unit 320 calculates the within-cluster sum of squared errors (WCSS) under different K values through the elbow method, plots the relationship curve between the K value and the within-cluster sum of squared errors, and determines the number of clusters;

[0077] First, obtain the dataset to be subjected to cluster analysis, ensure that the data has been cleaned, preprocessed, etc., removing factors such as noise and missing values that may affect the clustering results. Create an empty list to store the within-cluster sum of squared errors (WCSS) values under different K values. Traverse different K values. For each K value, run the K-Means clustering algorithm to calculate the WCSS under the current K value. Use the K value as the abscissa and the WCSS as the ordinate to plot a line graph, that is, the elbow graph;

[0078] Observe the elbow graph and find the inflection point of the curve, that is, the "elbow point". The characteristic of this point is that before it, as the K value increases, the WCSS decreases relatively quickly. After it, as the K value increases, the rate of decrease of the WCSS significantly slows down. The K value corresponding to this point is the optimal number of clusters determined by the elbow method.

[0079] In order to better dynamically adjust the advertising marketing strategy, among them, the evaluation and feedback module 400 conducts a comparative analysis, compares the data of this advertising marketing activity with that of previous activities or the industry average level, finds out the advantages and gaps, and dynamically adjusts the advertising marketing strategy;

[0080] The customer response rate refers to the ratio of the number of customers who respond to an advertising and marketing campaign to the total number of customers who receive the marketing message. The calculation formula is: Customer Response Rate = (Number of Responding Customers ÷ Total Number of Reached Customers) × 100%. Responses can include behaviors such as clicking on an advertisement, replying to an email, making a consultation call, participating in online or offline activities, etc.;

[0081] The purchase conversion rate refers to the ratio of the number of customers who complete a purchase within a certain period of time to the number of customers who participate in the marketing campaign or visit the sales page. The calculation formula is: Purchase Conversion Rate = (Number of Purchasing Customers ÷ Number of Customers Participating in the Activity or Visiting the Page) × 100%;

[0082] The customer retention rate refers to the ratio of the number of customers who continue to use the product or service within a certain time period to the initial number of customers. Common ones include monthly retention rate, quarterly retention rate, etc. The calculation formula is: Customer Retention Rate = (Number of Retained Customers at the End ÷ Number of Customers at the Beginning) × 100%;

[0083] Adjust the strategy according to the evaluation results:

[0084] Low customer response rate: If the advertisement content fails to attract customers, the advertisement creativity can be redesigned to highlight the unique selling points and value of the product or service, and more attractive pictures, videos or copy can be used. If the advertising placement channels are not accurate, the advertising placement channels can be re-evaluated, and channels that are more in line with the characteristics of the target customer group can be selected, such as switching from social media platforms to industry vertical websites;

[0085] Low purchase conversion rate: Simplify the purchase process, reduce the number of operation steps for customers during the purchase process, improve the loading speed of the website or APP, ensure the safety and convenience of the payment link, and let customers understand the product or service more comprehensively and enhance the purchase confidence by adding product detail pictures, product usage videos, customer reviews, etc.;

[0086] Low customer retention rate: Strengthen after-sales customer care, improve customer satisfaction and loyalty by means of regular follow-up visits, sending care text messages or emails, providing exclusive membership services, etc., and conduct personalized marketing recommendations according to the customer's purchase history and preferences, and push product or service information that meets the customer's needs to improve customer participation and retention rate.

[0087] In order to better process the collected data, among them, the data acquisition module 100 checks for missing values, duplicate values and outliers in the data, and processes the data using the mean filling method;

[0088] Check for missing values, duplicate values, and outliers in the data and perform corresponding processing. For missing values, methods such as deleting missing value records, filling with the mean, or filling based on the information of similar customers can be selected according to the data characteristics. For duplicate values, only one record needs to be retained. For outliers, it is necessary to determine whether they are real data. If they are incorrect data, they should be corrected or deleted.

[0089] In summary, the working principle of this solution is as follows:

[0090] In the advertising and marketing system for after-sales data statistics, the basic information, purchase information, after-sales behavior data, and customer feedback data of customers are collected through the data collection module 100, and the data is processed. The intelligent analysis module 200 extracts features from the processed data. Taking the number of complaints and the interval between the last purchase time and the current time as input features, and the probability of customer churn as the output feature, a recurrent neural network model for predicting the probability of customer churn is established to realize predicting the probability of customer churn based on the after-sales data of old customers. The precision marketing module 300 takes the probability of customer churn predicted by the intelligent analysis module 200 as the horizontal axis and the customer value as the vertical axis to establish a two-dimensional matrix of probability division, and divides customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers, and low-value high-churn customers. The K-Means algorithm is used to cluster the customers within the quadrants, and corresponding advertising marketing strategies are formulated. Different advertising marketing strategies are adopted for different types of customers to improve the pertinence, maximize the recovery of customers, and reduce the losses of the enterprise. The evaluation and feedback module 400 takes the customer response rate, purchase conversion rate, and customer retention rate as evaluation indicators, regularly evaluates the effect of the advertising and marketing activities, dynamically adjusts the advertising marketing strategies formulated by the precision marketing module 300 according to the evaluation results, continuously optimizes the advertising marketing strategies, and improves the resource utilization rate.

[0091] The above shows and describes the basic principle, main features, and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An advertising and marketing system for after-sales data statistics, characterized by: It includes a data collection module (100), an intelligent analysis module (200), a precision marketing module (300) and an evaluation feedback module (400); The data collection module (100) collects basic customer information, purchase information, after-sales behavior data, and customer feedback data from the enterprise's after-sales system and customer service records, cleans and annotates the collected data, and transmits the processed data to the intelligent analysis module (200); The intelligent analysis module (200) uses a correlation analysis method to extract features from the data collected by the data collection module (100), encodes and normalizes the extracted features, uses the number of complaints and the interval between the last purchase time and the current time as input features, and uses the predicted customer churn probability as output features to establish a recurrent neural network model for predicting the customer churn probability; The precision marketing module (300) uses the customer churn probability predicted by the intelligent analysis module (200) as the horizontal axis and the customer value as the vertical axis to establish a two-dimensional matrix of probability division, divides customers into high-value low churn customers, high-value high churn customers, low-value low churn customers and low-value high churn customers, clusters customers within the range using the K-Means algorithm, and formulates corresponding advertising marketing strategies; The evaluation feedback module (400) uses customer response rate, purchase conversion rate and customer retention rate as evaluation indicators to regularly evaluate the effectiveness of advertising and marketing activities, and dynamically adjusts the advertising and marketing strategy formulated by the precision marketing module (300) based on the evaluation results.

2. The advertising and marketing system for after-sales data statistics according to claim 1, characterized in that: The intelligent analysis module (200) comprises a feature extraction unit (210) and a model building unit (220); The feature extraction unit (210) extracts features from the collected data using a correlation analysis method, and arranges the extracted feature data in a time series; The model building unit (220) uses the ReLU function as the activation function of the hidden layer and the Softmax function as the activation function of the output layer to build a neural network model for predicting the probability of customer churn, and uses the Adam optimizer to train the model.

3. The advertising and marketing system for after-sales data statistics according to claim 2 is characterized by: The feature extraction unit (210) converts the categorical features into numerical features by means of a label encoding method, and maps the feature values ​​into a range of 0 to 1 by means of a Z-score normalization method.

4. The advertising and marketing system for after-sales data statistics according to claim 2 is characterized by: The model building unit (220) selects a multi-classification cross entropy loss function as the loss function of the model, which is used to measure the difference between the model prediction result and the actual situation. The formula is: ； in, is the loss function value, is the number of categories, is the true category label, For the prediction of The probability of the class.

5. The advertising and marketing system for after-sales data statistics according to claim 2 is characterized by: The model building unit (220) uses a grid search method to optimize the hyperparameters of the recurrent neural network model for predicting the probability of customer churn, combines the values ​​of each hyperparameter to form a grid-shaped parameter space, performs model training and evaluation on each group of parameter combinations in the parameter space, and selects the parameter combination with the best performance on the validation set as the optimal hyperparameter.

6. The advertising and marketing system for after-sales data statistics according to claim 1 is characterized by: The precision marketing module (300) includes a probability division unit (310) and a strategy formulation unit (320); The probability division unit (310) uses a matrix analysis method to divide customers according to customer churn probability and customer value, and divides customers into high-value low-churn customers, high-value high-churn customers, low-value low-churn customers and low-value high-churn customers; The strategy formulation unit (320) clusters the customers in the quadrant using the K-Means algorithm, takes the customers' consumption amount and consumption frequency as clustering features, and formulates corresponding advertising marketing strategies for the customers in the same cluster.

7. The advertising and marketing system for after-sales data statistics according to claim 6 is characterized by: The probability division unit (310) uses the RFM model to establish a customer value assessment system, and determines the customer value by calculating the customer's most recent consumption, consumption frequency and consumption amount.

8. The advertising and marketing system for after-sales data statistics according to claim 6 is characterized by: The strategy formulation unit (320) calculates the sum of squared errors within the cluster under different K values ​​by the elbow method, draws a relationship curve between the K value and the sum of squared errors within the cluster, and determines the number of clusters.

9. The advertising and marketing system for after-sales data statistics according to claim 1, characterized in that: The evaluation feedback module (400) compares the data of the current advertising and marketing campaign with previous campaigns or the industry average through comparative analysis, finds out the advantages and gaps, and dynamically adjusts the advertising and marketing strategy.

10. The advertising and marketing system for after-sales data statistics according to claim 1, characterized in that: The data acquisition module (100) checks missing values, duplicate values ​​and abnormal values ​​in the data, and processes the data using a mean filling method.

Images