Customer behavior trend analysis system based on terminal big data

Abstract

The invention discloses a client behavior trend analysis system based on terminal big data, and relates to the technical field of big data analysis, and the client behavior trend analysis system comprises a data collection module, a terminal health degree evaluation module, an intelligent prediction recommendation module, and a user experience improvement module. The health degree of the terminal is evaluated by analyzing the use data of the terminal, the problems of the terminal, namely insufficient service life of a battery, insufficient storage space and application performance reduction, can be found in time, timely maintenance and replacement suggestions are provided for a user, and meanwhile, the behavior trend of a customer is intelligently predicted based on the health degree evaluation result of the terminal. According to the embodiment of the invention, personalized services and recommendations can be provided for the user, personalized requirements of the user can be met, and suitable applications, services and preferential activities can be recommended for the user according to use habits and preferences of the user, so that the user can obtain required resources more conveniently, and the comfort level when the user uses the terminal equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a customer behavior trend analysis system based on terminal big data. Background Art

[0002] With the rapid development of mobile Internet, the popularity of smart terminals is increasing, and people's lives and work are increasingly dependent on various terminal devices. The big data generated by terminal devices contains rich information, which is of great significance for in-depth understanding of customer behavior, predicting behavior trends and providing personalized services. In this context, a customer behavior trend analysis system based on terminal big data came into being. It aims to provide decision support for enterprises and improve user experience and loyalty by exploring customers' potential needs and behavior patterns through in-depth analysis of terminal data.

[0003] However, the existing customer behavior trend analysis system only focuses on the status of a single hardware component in the terminal health assessment part, and only assesses the battery health through the number of battery charge and discharge times, ignoring the impact of other important hardware (such as storage space) and software (such as application performance) on the overall health of the terminal. It usually performs simple correlation analysis based on user behavior data and fails to deeply explore the deep-seated reasons behind user behavior, resulting in inaccurate recommendation results. Furthermore, most terminal health assessment systems lack a linkage mechanism with user behavior, making it difficult to dynamically adjust the assessment model according to the actual user usage, causing the assessment results to lag behind the actual situation, ignoring the continuous optimization of user experience, and failing to improve user satisfaction and loyalty through personalized services.

[0004] To sum up, the existing terminal health assessment and intelligent prediction and recommendation have the problems of single assessment dimension, low prediction and recommendation accuracy, and insufficient dynamic adjustment capability, which limits the accuracy of the system's terminal health assessment and also affects the effectiveness of the intelligent recommendation service. Therefore, there is an urgent need for a new customer behavior trend analysis system based on terminal big data that can comprehensively evaluate the terminal health and make intelligent predictions and recommendations on this basis, thereby improving user experience and loyalty. Summary of the invention

[0005] The purpose of the present invention is to make up for the shortcomings of the prior art and to provide a customer behavior trend analysis system based on terminal big data. The system realizes all-round health monitoring and intelligent recommendation services for smart terminal devices by integrating a behavior trend prediction unit and a recommendation generation unit. In combination with the user's historical behavior patterns, a high-precision terminal health assessment model is constructed, so that the system can intelligently predict the user's future usage needs, and tailor services and application recommendations for users through the recommendation generation unit, thereby enhancing the personalization and pertinence of the service, and providing users with a more convenient, efficient and intelligent smart terminal usage experience.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a customer behavior trend analysis system based on terminal big data, the components of which include: a data collection module, a terminal health assessment module, an intelligent prediction and recommendation module, and a user experience improvement module;

[0007] The data collection module is used to collect terminal usage data, and the terminal usage data includes battery life, storage space, and application performance;

[0008] The terminal health evaluation module is connected to the data acquisition module, and is used to analyze the terminal usage data, evaluate the terminal health, and output the evaluation result of the terminal health;

[0009] The intelligent prediction and recommendation module is connected to the terminal health evaluation module, intelligently predicts customer behavior trends based on the evaluation results and outputs the prediction results of customer behavior;

[0010] The user experience improvement module is connected to the intelligent prediction and recommendation module to collect user feedback on the recommended service and adjust the recommendation strategy according to the prediction results;

[0011] The terminal health evaluation module is composed of a data preprocessing unit and a health evaluation unit, which cleans, organizes and analyzes the collected data and calculates the health score of the terminal;

[0012] A data preprocessing unit, used to clean and process the terminal usage data to remove noise and abnormal data;

[0013] A health evaluation unit, connected to the data preprocessing unit, for analyzing the processed terminal usage data using a health evaluation model to evaluate the terminal health;

[0014] The intelligent prediction and recommendation module consists of a behavior trend prediction unit and a recommendation generation unit. It extracts the user's past behavior patterns from the terminal usage logs and uses historical data to train and verify the evaluation model to predict the user's future behavior trends. It specifically includes:

[0015] A behavior trend prediction unit, which predicts customer behavior trends based on the terminal health evaluation result of the health evaluation unit, combined with historical data and user portraits, wherein the user portraits include customer terminal characteristics, home service subscription characteristics, high-precision location characteristics, customer channel preferences, and customer login information;

[0016] The recommendation generating unit is connected to the behavior trend predicting unit and is used to generate personalized recommendation content according to the prediction result of the behavior trend predicting unit on the customer behavior trend, and the recommendation content includes application recommendation, service recommendation and preferential activity recommendation.

[0017] Furthermore, the data collection module is responsible for collecting various operation data of the terminal device, including:

[0018] Battery life detection: Through the battery management of terminal equipment, the number of battery charge and discharge times, current power level, and battery health index are collected;

[0019] Storage space monitoring: Regularly check the storage space usage of terminal devices, including remaining space and percentage of used space;

[0020] Application performance tracking: Records performance indicators such as application startup time, CPU and memory consumption during runtime, and crash frequency.

[0021] Furthermore, the health evaluation unit uses the battery life, storage space and application performance usage data in the terminal as the data features of the input layer, which are recorded as x1, x2, x3, ..., x n , the terminal health score H is defined as the output result, and a multi-layer neural network is used to build a health assessment model. The health assessment model structure includes an input layer, a hidden layer, and an output layer.

[0022] Furthermore, the hidden layer of the health assessment model is divided into a first hidden layer and a second hidden layer, wherein:

[0023] The first hidden layer normalizes the input terminal usage data and uses the processed feature data x′ i The output of the first hidden layer is obtained by calculating the activation function, that is, Among them, h j1 is the output of the j1th node in the first hidden layer, w ij1 is the weight from the input layer to the first hidden layer, b j1 is the bias of the first hidden layer, n is the total number of features in the terminal health assessment;

[0024] The output of the second hidden layer is also calculated by the activation function, that is, where h j2 is the output of the j2th node in the second hidden layer, w j1j2 is the weight from the first hidden layer to the second hidden layer, b j2 is the bias of the second hidden layer, and m1 represents the number of nodes in the first hidden layer.

[0025] Furthermore, the output of the output layer of the health assessment model is the terminal health score H, which is calculated by a linear function, and the calculation formula is: where w j2h is the weight from the hidden layer to the output layer, b h is the bias of the output layer, and m2 represents the number of nodes in the second hidden layer.

[0026] Furthermore, the health assessment model uses historical data to train the health assessment model. During the training process, the weights and biases are adjusted to minimize the prediction error. The mean square error loss function is used, and the calculation formula is: Where N is the number of samples, H i Score the actual terminal health, The terminal health score predicted by the health assessment model is evaluated using the test set. The evaluation indicators include accuracy and recall. The accuracy is used to measure the correct proportion of the health assessment model predictions, and the recall is used to measure the proportion of terminals that the health assessment model can correctly identify with low actual health. These evaluation indicators are used to evaluate the performance and generalization ability of the health assessment model.

[0027] Furthermore, the behavior trend prediction unit fuses the terminal health assessment result, historical data and user portrait, and takes the terminal health score H as the terminal health assessment result, the historical data as D, and the user portrait feature vector as P = [p1, p2, p3, p4, p5], where p1 represents the customer terminal feature, p2 represents the home service subscription feature, p3 represents the high-precision location feature, p4 represents the customer channel preference, and p5 represents the customer login information. Feature extraction is performed on the fused data to obtain the reduced dimension feature vector F = [f1, f2, ..., f m ], use the neural network model for structural training, take the feature vector F as the input feature, output the predicted customer behavior trend T through the activation function, and calculate the hidden layer: where h j is the output of the jth node in the hidden layer, w ij is the weight from the input layer to the hidden layer, b j is the bias of the hidden layer, σ is the activation function, f i is the i-th element in the feature vector F, m is the total number of feature vectors F, and the output layer is calculated as: Where T is the predicted customer behavior trend vector, w jh is the weight from the hidden layer to the output layer, h j is the output of the jth node in the hidden layer, b h is the bias of the output layer.

[0028] Furthermore, the recommendation generation unit calculates the correlation between the user and the behavior trend based on the customer behavior trend vector T predicted by the behavior trend prediction unit, combined with the user portrait and historical recommendation data, that is, Among them, u represents the user, T is the predicted customer behavior trend vector, and I uT represents the set of items associated with the user and the behavior trend, f ui Represents the feature vector of the user's response to the project, generated based on the user profile and historical behavior data, T i is the part of the behavior trend vector related to the project, and the user’s interest in the project is predicted based on the correlation between the user and the behavior trend, that is, in, represents the average interest of user u, rel(u,T) ​​is the correlation calculation function between user u and the predicted customer behavior trend vector T, N(u) represents the set of users similar to user u, sim(u,v) represents the similarity between user u and user v, r vi represents the interest of user v in item i, represents the average interest of user u, Represents the average interest of user v. Recommended content is generated based on the predicted interest, that is, items with high predicted interest are selected as recommended content, including application recommendations, service recommendations, and promotion recommendations.

[0029] Furthermore, the user experience improvement module includes:

[0030] A personalized service providing unit, used to provide personalized services to users according to the prediction results of the intelligent prediction and recommendation module;

[0031] The loyalty enhancement unit is connected to the personalized service providing unit and is used to enhance user loyalty by providing high-quality services and experiences.

[0032] Compared with the existing technology, this customer behavior trend analysis system based on terminal big data has the following beneficial effects:

[0033] 1. The present invention evaluates the health of the terminal by analyzing the terminal usage data, and can promptly discover the problems existing in the terminal, namely, insufficient battery life, tight storage space, and reduced application performance, and provide users with timely maintenance and replacement suggestions. At the same time, based on the terminal health evaluation results, it can intelligently predict customer behavior trends and provide users with personalized services and recommendations to meet the personalized needs of users. According to the user's usage habits and preferences, it recommends suitable applications, services and preferential activities to users, so that users can more conveniently obtain the required resources, thereby improving the comfort and convenience of users when using terminal devices.

[0034] 2. The present invention realizes a comprehensive assessment and optimization of the health status of smart terminal devices by constructing an advanced terminal health assessment model, accurately identifies and intervenes in advance the factors that affect the health of the device, effectively extends the service life of the smart terminal, and accurately pushes relevant services and applications based on the user's usage habits and demand forecasts, which not only reduces the cost of users finding the required information by themselves, but also improves the efficiency of information utilization and promotes the healthy development of information consumption.

[0035] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0037] Figure 1 This is an operation flow chart of a customer behavior trend analysis system based on terminal big data;

[0038] Figure 2 This is a module organization diagram of a customer behavior trend analysis system based on terminal big data. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. 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.

[0040] Embodiment 1

[0041] This embodiment describes a specific application process of a customer behavior trend analysis system based on terminal big data in a mobile operator to improve user satisfaction and loyalty. The customer behavior trend analysis system of the present invention is deployed. The operator has a large number of smart terminal users and collects and analyzes the usage data of user terminals through the system to optimize service quality and recommendation strategies.

[0042] In terms of the data acquisition module, the system collects usage data of the user terminal in real time. For example, for battery life, the system obtains information such as the number of battery charge and discharge times, the current power level, and the battery health index through the battery management system of the terminal device. Assuming that the user's mobile phone battery has been charged and discharged 500 times, the current power level is 30%, and the battery health index is 80%. For storage space, the system regularly checks the storage space usage of the terminal device, including the remaining space and the proportion of used space. Assuming that the total storage space of the mobile phone is 128GB, the used space is 64GB, the remaining space is 64GB, and the proportion of used space is 50%. For application performance, the system records performance indicators such as the application startup time, CPU and memory consumption during runtime, and crash frequency. Assuming that the average startup time of the application is 3 seconds, the average CPU consumption during runtime is 20%, the average memory consumption is 500MB, and the crash frequency is once a week, the terminal health assessment module receives these data and performs terminal health assessment. Taking battery life assessment as an example, the health assessment model is used to process the data first, and normalize the battery health index and other data. Assuming that the normalized battery health index ranges from 0 to 1, 1 indicates the best health, the hidden layer calculation formula is: Assume that the weights w of the first hidden layer are ij1 and b j1 The biases are [0.5, 0.3, 0.2] and [10, 5, 20] respectively, then the output h of the first hidden layer is j1 is max(0,0.5×0.8+10)=10.4. Similarly, the output of the second hidden layer further extracts and integrates the features of the data, allowing the model to more accurately evaluate the health of the terminal, that is,

[0043] The weights w of the second hidden layer j1j2 and bias b j2 and 0.4, 0.6, 0.3] and [8, 6, 15] respectively, then the output of the second hidden layer is max(0, 0.4×10.4+8)=12.16, and the terminal health score is calculated through the output layer. That is, 0.7×12.16+20=28.512. Then, the intelligent prediction and recommendation module predicts customer behavior trends based on the terminal health evaluation results. Assuming that the terminal health score is low, the behavior trend prediction unit fuses the terminal health evaluation results, historical data (such as the user's past phone replacement frequency, the type of application used) and user portraits through data fusion, and adopts the algorithm of the behavior trend prediction unit to extract features and train models. The weight w from the input layer to the hidden layer ij and bias b jare [0.6, 0.4, 0.3] and [5, 3, 10] respectively, then the output h of the hidden layer j =σ(0.6×0.8+5)=

[0044] σ(5.48), the predicted customer behavior trend vector T is That is, T = 20.7, where w jh is the weight from the hidden layer to the output layer, h j is the output of the jth node in the hidden layer, b h is the bias of the output layer, k represents the total number of nodes when calculating from the hidden layer to the output layer. The purpose of calculating the user behavior trend value is to predict the customer's future behavior trend, so as to provide a basis for personalized recommendation. The value reflects the strength of the customer's behavior trend. For example, a value of 20.7 indicates that the customer has a certain behavior trend. If the value is too large, it means that the customer has a strong demand for changing the phone or is interested in some new services; if the value is too small, it means that the customer's behavior trend is relatively stable without obvious changes.

[0045] The recommendation generation unit generates personalized recommendation content based on predicted customer behavior trends. For example, based on the user's behavior trends and user portraits, it is predicted that the user may have the need to replace the terminal and has a high interest in high-performance mobile phones and high-speed networks. Therefore, the recommendation generation unit will recommend the user to purchase the latest high-end models and recommend upgrading to a higher-speed package. At the same time, based on the characteristics of users' frequent activities in commercial areas, relevant commercial discount activities are recommended. In terms of the user experience improvement module, when the user receives these personalized recommendations, if the user chooses to replace the terminal, the system will provide the user with relevant purchasing channels and discount information, and remind the user to back up data and transfer information in time. If the user chooses to upgrade the package, the system will assist the user to complete the package upgrade operation and ensure the smooth progress of the upgrade process. By using this system, the operator can understand the user's needs and behavior trends more accurately, thereby providing more personalized services and recommendations. For example, for users with low terminal health, the system will promptly remind users to perform maintenance or replace terminals to avoid terminal problems affecting user experience. For users who have the potential to change their phones, suitable models and promotions can be recommended in advance to increase users' willingness to buy. At the same time, through the analysis of user behavior trends, operators can optimize business strategies. For example, they can launch more suitable data packages for users with large data usage, and launch relevant regional promotions for users who often move in specific areas, thereby improving user satisfaction and loyalty.

[0046] In addition, through the in-depth analysis of terminal big data by the data analysis module, operators can explore potential user needs and behavior patterns and provide strong support for market decisions. For example, if it is found that users have a high demand for video applications, operators can increase network optimization efforts in the area to improve the smoothness of video playback, and launch related video packages and promotions to meet user needs. At the same time, the terminal health assessment results and customer behavior trend prediction results are displayed to the operator's managers in a visual way, so that they can intuitively understand the user's terminal health status and behavior trends, so as to better make decisions and optimize business strategies.

[0047] In summary, through the implementation of the customer behavior trend analysis system based on terminal big data described in this patent, the mobile operator can effectively improve user experience, optimize business strategies, enhance market competitiveness, and achieve sustainable development.

[0048] Embodiment 2

[0049] This embodiment provides an e-commerce company that uses the customer behavior trend analysis system to improve user shopping experience, precision marketing, and increase user retention rate.

[0050] In terms of the data collection module, the system collects various data on user terminals, including but not limited to users' browsing history, purchase records, search keywords, performance parameters of terminal devices, etc. For example, a user has browsed pages of multiple electronic products in the past month, purchased some household items, and frequently searched for keywords such as "smart watches". Data such as the battery life, storage space, and application performance of his terminal devices are also collected in real time.

[0051] The terminal health assessment module performs terminal health assessment based on the collected data. Taking battery life as an example, assuming that the battery health index of a user's terminal is, after normalization, combined with other relevant data, it is calculated through the evaluation model, and the terminal health score is calculated through the output layer. The intelligent prediction and recommendation module predicts customer behavior trends based on the terminal health assessment results and other relevant data. For example, based on the user's browsing history and purchase history, it is found that the user has a high degree of attention to electronic products and has frequently browsed smart watch related pages recently. Combined with user portraits (such as age, gender, region, etc.), it is predicted that the user may have the intention to purchase smart watches, and feature extraction and model training are performed. The recommendation generation unit generates personalized recommendation content based on the predicted customer behavior trends. Based on the above prediction, the recommendation generation unit will recommend related smart watch products to the user, and recommend some supporting electronic products or peripheral accessories based on the user's browsing history and preferences. At the same time, combined with the user's terminal health, if it is found that the terminal storage space is insufficient, some cloud storage services or cleaning tools will be recommended.

[0052] In terms of the user experience improvement module, the personalized service provision unit customizes exclusive services for users by analyzing the prediction results of the intelligent prediction and recommendation module, combining user portraits and historical behavior data. For example, if it is predicted that the user has travel needs, the user is recommended suitable travel methods, tourist attractions and hotels based on the user's travel preferences (such as whether to drive or take public transportation, preferred travel destination types, etc.), consumption habits (budget range, etc.) and current location information. The loyalty improvement unit improves user loyalty by providing users with high-quality and personalized service experience. Specific methods include providing exclusive discounts for old users, such as issuing discount coupons and doubling points for long-term and active users; providing users with priority services, such as giving priority in customer service consultation and after-sales processing; holding member-exclusive activities to enhance users' sense of identity and dependence on products or services, thereby improving user loyalty. When users receive these personalized recommendations, if they are interested in a certain smart watch, the system will provide detailed product information, user reviews and purchase suggestions. If the user's terminal storage space is insufficient, the system will remind the user to clean up unnecessary files or provide solutions for expanding storage.

[0053] In actual applications, by using this system, the e-commerce company can understand users' needs and behavior trends more accurately, and push relevant product information to users with purchasing intentions in a timely manner to improve purchase conversion rates; for potential users, their purchasing desire is stimulated through personalized recommendations. At the same time, according to user behavior trends, the website layout and recommendation algorithm are optimized to improve the user's shopping experience. In addition, through the in-depth analysis of terminal big data by the data analysis module, the e-commerce company can discover users' potential needs and market trends. For example, if a certain type of electronic product is found to be in high demand among users in a specific region or age group, product strategies and marketing activities can be adjusted in a targeted manner, and terminal health assessment results, customer behavior trend prediction results, etc. can be displayed to managers in a visual way to help them make more informed decisions, such as optimizing product recommendation strategies and adjusting inventory management.

[0054] To sum up, through the implementation of the customer behavior trend analysis system based on terminal big data described in this patent, the e-commerce company can better meet user needs, improve user satisfaction and retention rate, and thus enhance market competitiveness.

[0055] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A customer behavior trend analysis system based on terminal big data, characterized in that: The components of the customer behavior trend analysis system include: data collection module, terminal health assessment module, intelligent prediction and recommendation module, and user experience improvement module; The data collection module is used to collect terminal usage data, and the terminal usage data includes battery life, storage space, and application performance; The terminal health evaluation module is connected to the data acquisition module, and is used to analyze the terminal usage data, evaluate the terminal health, and output the evaluation result of the terminal health; The intelligent prediction and recommendation module is connected to the terminal health evaluation module, intelligently predicts customer behavior trends based on the evaluation results, and outputs the prediction results of customer behavior; The user experience improvement module is connected to the intelligent prediction and recommendation module to collect user feedback on the recommended service and adjust the recommendation strategy according to the prediction results; The terminal health evaluation module is composed of a data preprocessing unit and a health evaluation unit, which cleans, organizes and analyzes the collected data and calculates the health score of the terminal; A data preprocessing unit, used to clean and process the terminal usage data to remove noise and abnormal data; A health evaluation unit, connected to the data preprocessing unit, for analyzing the processed terminal usage data using a health evaluation model to evaluate the terminal health; The intelligent prediction and recommendation module consists of a behavior trend prediction unit and a recommendation generation unit. It extracts the user's past behavior patterns from the terminal usage logs and uses historical data to train and verify the evaluation model to predict the user's future behavior trends. It specifically includes: A behavior trend prediction unit, which predicts customer behavior trends based on the terminal health evaluation result of the health evaluation unit, combined with historical data and user portraits, wherein the user portraits include customer terminal characteristics, home service subscription characteristics, high-precision location characteristics, customer channel preferences, and customer login information; The recommendation generating unit is connected to the behavior trend predicting unit and is used to generate personalized recommendation content according to the prediction result of the behavior trend predicting unit on the customer behavior trend, and the recommendation content includes application recommendation, service recommendation and preferential activity recommendation.

2. According to claim 1, a customer behavior trend analysis system based on terminal big data is characterized in that: The data acquisition module is responsible for collecting various operation data of the terminal equipment, including: Battery life detection: Through the battery management of terminal equipment, the number of battery charge and discharge times, current power level, and battery health index are collected; Storage space monitoring: Regularly check the storage space usage of terminal devices, including remaining space and percentage of used space; Application performance tracking: Records performance indicators such as application startup time, CPU and memory consumption during runtime, and crash frequency.

3. The customer behavior trend analysis system based on terminal big data according to claim 1 is characterized in that: The health evaluation unit uses the battery life, storage space and application performance usage data in the terminal as the data features of the input layer, denoted as x1, x2, x3, ..., x n , the terminal health score H is defined as the output result, and a multi-layer neural network is used to build a health assessment model. The health assessment model structure includes an input layer, a hidden layer, and an output layer.

4. The customer behavior trend analysis system based on terminal big data according to claim 3 is characterized in that: The hidden layer of the health assessment model is divided into a first hidden layer and a second hidden layer, wherein the first hidden layer normalizes the input terminal usage data and uses the processed feature data x′ i , the output of the first hidden layer is obtained by calculating the activation function; Right now Among them, h j1 is the output of the j1th node in the first hidden layer, w ij1 is the weight from the input layer to the first hidden layer, b j1 is the bias of the first hidden layer, n is the total number of features in the terminal health assessment; The output of the second hidden layer is also calculated by the activation function; Right now Among them, h j2 is the output of the j2th node in the second hidden layer, w j1j2 is the weight from the first hidden layer to the second hidden layer, b j2 is the bias of the second hidden layer, and m1 represents the number of nodes in the first hidden layer.

5. The customer behavior trend analysis system based on terminal big data according to claim 4 is characterized in that: The output of the output layer of the health assessment model is the terminal health score H, which is calculated by a linear function. The calculation formula is: Among them, w j2h is the weight from the hidden layer to the output layer, b h is the bias of the output layer, and m2 represents the number of nodes in the second hidden layer.

6. The customer behavior trend analysis system based on terminal big data according to claim 5 is characterized in that: The health assessment model uses historical data to train the health assessment model. During the training process, the weights and biases are adjusted to minimize the prediction error. The mean square error loss function is used, and the calculation formula is: Where N is the number of samples, H i Score the actual terminal health, The terminal health score predicted by the health assessment model is evaluated using the test set. The evaluation indicators include accuracy and recall. The accuracy is used to measure the correct proportion of the health assessment model predictions, and the recall is used to measure the proportion of terminals that the health assessment model can correctly identify with low actual health. These evaluation indicators are used to evaluate the performance and generalization ability of the health assessment model.

7. The customer behavior trend analysis system based on terminal big data according to claim 6 is characterized in that: The behavior trend prediction unit fuses the terminal health evaluation result, historical data and user portrait, and uses the terminal health score H as the terminal health evaluation result, the historical data as D, and the user portrait feature vector as P = [p1, p2, p3, p4, p5]; wherein p1 represents the customer terminal feature, p2 represents the home service subscription feature, p3 represents the high-precision location feature, p4 represents the customer channel preference, and p5 represents the customer login information; the fused data is subjected to feature extraction to obtain the reduced dimension feature vector F = [f1, f2, ..., f m ], using the neural network model for structural training, taking the feature vector F as the input feature, and outputting the predicted customer behavior trend T through the activation function, its hidden layer calculation: Among them, h j is the output of the jth node in the hidden layer, w ij is the weight from the input layer to the hidden layer, b j is the bias of the hidden layer, σ is the activation function, f i is the i-th element in the feature vector F, m is the total number of feature vectors F, and the output layer is calculated as: Where T is the predicted customer behavior trend vector, w jh is the weight from the hidden layer to the output layer, h j is the output of the jth node in the hidden layer, b h is the bias of the output layer, and k represents the total number of nodes from the hidden layer to the output layer.

8. The customer behavior trend analysis system based on terminal big data according to claim 7 is characterized in that: The recommendation generation unit calculates the correlation between the user and the behavior trend based on the customer behavior trend vector T predicted by the behavior trend prediction unit, combined with the user portrait and historical recommendation data; Among them, u represents the user, T is the predicted customer behavior trend vector, and I uT represents the set of items associated with the user and the behavior trend, f ui Represents the feature vector of the user's response to the project, generated based on the user profile and historical behavior data, T i is the part of the behavior trend vector related to the project, and the user’s interest in the project is predicted based on the correlation between the user and the behavior trend, that is, in, represents the average interest of user u, rel(u,T is the correlation calculation function between user u and the predicted customer behavior trend vector T, N(u) represents the set of users similar to user u, sim(u,v) represents the similarity between user u and user v, r vi represents the interest of user v in item i, represents the average interest of user u, Represents the average interest of user v. Recommended content is generated based on the predicted interest, that is, items with high predicted interest are selected as recommended content, including application recommendations, service recommendations, and promotion recommendations.

9. The customer behavior trend analysis system based on terminal big data according to claim 1 is characterized in that: The user experience improvement module includes: A personalized service providing unit, used to provide personalized services to users according to the prediction results of the intelligent prediction and recommendation module; The loyalty enhancement unit is connected to the personalized service providing unit and is used to enhance user loyalty by providing high-quality services and experiences.

Images