User full-cycle journey experience system and medium

Abstract

The invention provides a user full-cycle journey experience system and a medium. The user full-cycle journey experience system comprises data acquisition: collecting multi-dimensional data of a user; demand prediction: predicting user demands by using deep learning; preliminary optimization: calculating optimization parameters, and adjusting experience; feedback acquisition: collecting multi-modal feedback data; experience detection: evaluating a current experience accurate value; and deep optimization: adjusting parameters according to the deviation, and optimizing experience. According to the method, dynamic prediction of user demands and continuous optimization of experience can be realized, and the satisfaction degree and loyalty of the user in a whole-cycle journey are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of user experience optimization, and more specifically, to a user full-cycle journey experience system and medium. Background Art

[0002] In today's digital age, users are increasingly interacting with various systems and products, and the importance of user experience is becoming increasingly prominent. Traditional user experience optimization methods rely primarily on user feedback and simple data analysis, which have many limitations. For example, user feedback is often delayed and cannot reflect changes in user needs in real time. Data analysis also tends to focus on a single dimension, such as user behavior data or demographic information, making it difficult to fully capture users' complex needs and psychological states. Furthermore, existing systems lack consideration of the user's full journey when optimizing the user experience, often focusing only on specific links, resulting in gaps in the user experience at different stages.

[0003] Existing user demand forecasting technologies are mostly based on simple statistical models or machine learning algorithms. These models lack accuracy and generalization capabilities when processing high-dimensional, complex data. Furthermore, user experience testing methods are relatively limited and cannot accurately assess users' real-world experiences in multimodal interaction environments. Optimization strategies lack dynamic adjustment mechanisms, making it difficult to quickly and effectively optimize based on real-time user feedback.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technology: it is impossible to obtain and analyze users' multi-dimensional data in a comprehensive and real-time manner, resulting in inaccurate predictions of user needs; user experience detection is not accurate enough and cannot cover multimodal interaction scenarios; the optimization strategy lacks dynamism and depth, and cannot achieve continuous optimization of the user's full-cycle journey experience. Summary of the Invention

[0005] The present invention provides a user full-cycle journey experience system and medium.

[0006] In a first aspect of the present invention, a user full-cycle journey experience system is provided, comprising: User data acquisition module, used to obtain multi-dimensional basic data of users; A user demand prediction module is used to input the multi-dimensional basic data, the user's real-time interactive behavior data and the system usage frequency into a user demand dynamic prediction model based on deep learning to output a multi-dimensional prediction vector of user demand; an experience optimization parameter calculation module, configured to calculate a first experience optimization parameter set using an adaptive adjustment algorithm based on the difference between the multi-dimensional prediction vector of the user demand and a preset user demand rating vector, so as to make preliminary adjustments to the user experience so that the user experience reaches a first optimized state; The system feedback data acquisition module is used to obtain multimodal feedback data generated by the system after the user operates in the system; A user experience detection module is used to input the multi-dimensional basic data and the multimodal feedback data into a user experience precision detection model that has been pre-trained with massive user data and integrated with a reinforcement learning mechanism, and output the precise multi-dimensional experience value of the user in the current experience state; The experience optimization parameter adjustment module is used to calculate a second experience optimization parameter set using a multi-objective optimization algorithm based on the deviation between the multi-dimensional experience precise value and the preset multi-dimensional experience rated value, so as to make the user experience reach the second optimized state and realize the continuous optimization of the user's full-cycle journey experience.

[0007] Furthermore, the user data acquisition module includes: Demographic information collection unit, used to collect the user's age, gender, region, educational background, occupation category and income level through the detailed information filled in by the user during registration and subsequent supplementary information; A user stratification model unit, configured to input the demographic information into a user stratification model based on cluster analysis, and divide the users into different segmented groups by combining the users' consumption behavior data and usage habit data; The demand preference prediction unit is used to use the Bayesian probability model to predict the user's demand preferences in the scenario based on the user's segmented group, the current usage scenario and time. The demand preferences include multiple dimensions such as demand preferences for functions, demand preferences for interface styles, and demand preferences for interaction methods.

[0008] Furthermore, the user data acquisition module further includes: The psychological characteristic data collection unit is used to extract psychological characteristic dimension data such as users' emotional tendencies, cognitive styles, motivation levels, and personality traits through natural language processing technology and sentiment analysis algorithms based on the psychological assessment questionnaires, behavioral tasks, and interaction records with intelligent customer service that users participate in in the system. A psychological profile model unit, configured to input the psychological characteristic data into a psychological profile model based on a deep neural network. The model adopts a long short-term memory network (LSTM) structure, which can capture the changing trend of the user's psychological characteristics over time and generate a psychological profile of the user; The user churn prediction unit is used to predict the probability of user churn and the possible churn time window using the Markov chain model based on the user's historical psychological state change trajectory and current usage behavior pattern if the user's psychological profile shows that he or she is anxious and has a low motivation level.

[0009] Furthermore, the user data acquisition module further includes: The interactive behavior data collection unit is used to collect multi-dimensional behavioral data such as user click behavior, sliding behavior, dwell time, operation path, and interaction frequency with system elements in various functional modules in real time through the system's built-in behavior monitoring module; A behavior pattern mining model unit is used to input the interactive behavior data into a behavior pattern mining model based on a graph neural network. The model can construct the user's behavior data into a behavior graph and mine the user's behavior pattern and preference path through graph convolution operations and node feature extraction; The function module optimization unit is used to optimize the operation process of a specific function module based on the business logic of the function module and the user's historical behavior data, using the Q-learning algorithm in reinforcement learning to improve the user's operation efficiency and experience satisfaction if the user's behavior pattern shows high frequency use of a specific function module and a relatively fixed path.

[0010] Furthermore, the user data acquisition module further includes: The usage time recording unit is used to accurately record the cumulative usage time, single usage time and the distribution of stay time in different functional modules after each user logs into the system through the system's time statistics module; A user value assessment model unit is used to combine the usage time data with the user's multi-dimensional basic data and interactive behavior data, and input the data into a user value assessment model based on a hybrid model. The model integrates a linear regression model and a nonlinear support vector machine model, comprehensively considering multiple dimensions such as the user's behavioral activity, loyalty, and consumption potential, to assess the user's comprehensive value level; The personalized service recommendation unit is used to provide users with customized service packages and recommended content based on the user's comprehensive value level, the current market environment and business goals, using dynamic pricing models and personalized recommendation algorithms to increase the user's usage time and consumption willingness.

[0011] Furthermore, the system feedback data acquisition module includes: The visual feedback data acquisition unit is used to capture the user's eye movement trajectory in real time through the embedded eye tracking device while the user is using the system, and record the user's gaze point, gaze duration, and scanning path of different interface elements; An interface visual perception model unit is used to combine the eye tracking data with the attribute information of the interface elements and input them into an interface visual perception model based on a convolutional neural network. The model adopts the Inception-ResNet architecture and learns the user's perception of the interface visual elements through multi-scale feature extraction and residual connections. The model outputs an interface color perception score, an element layout satisfaction score, and an icon recognition score. The auditory feedback data acquisition unit is used to collect the user's voice signal through the microphone array during the voice interaction between the user and the system, and record the sound effect signal output by the system; An auditory feedback analysis model unit is used to input the speech signal and sound effect signal into an auditory feedback analysis model based on deep speech recognition and audio feature extraction. The model uses a WaveNet architecture to generate high-quality speech features, and combines audio features such as Mel-frequency cepstral coefficients (MFCCs) and spectral centroids, using a support vector machine (SVM) classifier and regression model to evaluate sound effect quality, voice prompt clarity, and background music preference. The operation feedback data collection unit is used to collect the operation response time, number of erroneous operations and completion degree of the operation process in real time through the system performance monitoring module; The operation process analysis model unit is used to input the operation feedback data into the operation process analysis model based on operation process analysis. The model adopts the time series analysis method combined with the user experience quantitative model to evaluate the response speed, accuracy and convenience of the operation.

[0012] Furthermore, the output formula of the user demand dynamic prediction model is:

[0013] in, Represents the multi-dimensional prediction vector of user needs, is the weight matrix of the model, is the activation function, is the user's multi-dimensional basic data matrix, is the user's real-time interactive behavior data matrix, and are the bias vector and output bias vector of the model respectively; The weight matrix and the bias vector 、 The adjustment formula is:

[0014]

[0015]

[0016] in, is the learning rate, represents the gradient operator, is the loss function, It is a multi-dimensional vector of users' actual needs.

[0017] Furthermore, the input formula of the user experience precision detection model is:

[0018] in, represents the input vector of the model, Provide users with multi-dimensional basic data. is the multimodal feedback data of the system, Real-time interactive behavior data of users in the system, As the data fusion function, the principal component analysis PCA and linear discriminant analysis LDA are combined to reduce the dimension and extract features of the input data.

[0019] Furthermore, the output formula of the user experience precision detection model is:

[0020] in, Indicates the precise value of the user's multi-dimensional experience in the current experience state. is the prediction function of the model, is the parameter set of the model. The model adopts the Q-learning algorithm in deep reinforcement learning combined with the structure of deep neural network to continuously optimize the model parameters through interaction with users.

[0021] In a second aspect of the present invention, a computer-readable storage medium is provided, comprising instructions, which, when executed on a computer, enable the computer to execute any one of the methods according to the first aspect.

[0022] The above-described embodiments of the present invention have at least the following beneficial effects: The user journey experience system of the present invention can accurately predict and dynamically adjust user needs. By leveraging deep learning models and multi-dimensional data input, the system can capture changes in user needs in real time, calculate optimization parameters accordingly, and make preliminary adjustments to the user experience, ensuring that users receive an experience that better suits their needs.

[0023] In addition, the system can also obtain multimodal feedback data and use a precise detection model that integrates reinforcement learning mechanisms to accurately evaluate the user's multi-dimensional experience value in the current experience state. Further, based on the deviation between the experience value and the preset rated value, a multi-objective optimization algorithm is used to calculate the deep optimization parameters to achieve continuous optimization of the user's full-cycle journey experience and improve user satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which: Figure 1 A schematic diagram of the structure of a user full-cycle journey experience system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0025] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0026] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0027] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0028] Reference below Figure 1 , Figure 1 This is a schematic diagram of the structure of a user full-cycle journey experience system provided by an embodiment of the present invention. Figure 1 As shown, a user full-cycle journey experience system 100 includes: User data acquisition module 101, used to obtain multi-dimensional basic data of users, said multi-dimensional basic data including demographic information, psychological characteristic data, interactive behavior data and usage time of users; The user demand prediction module 102 is configured to input the multi-dimensional basic data, the user's real-time interactive behavior data, and the system usage frequency into a user demand dynamic prediction model based on deep learning, and output a multi-dimensional prediction vector of user demand through the model's neural network layer and activation function calculations; The experience optimization parameter calculation module 103 is configured to calculate a first experience optimization parameter set using an adaptive adjustment algorithm based on the difference between the multi-dimensional prediction vector of user needs and a preset user demand rating vector, thereby performing preliminary adjustments to the user experience to achieve a first optimized state. The first optimized state is defined as a state where, after the preliminary adjustments, user experience indicators such as efficiency and satisfaction in key operational links are significantly improved, and the user experience with primary functionalities more closely matches their expected needs. The adaptive adjustment algorithm may be based on the principles of an adaptive neural-fuzzy inference system (ANFIS), taking the difference between the multi-dimensional prediction vector of user needs and the preset user demand rating vector as input. Through the ANFIS's fuzzification, fuzzy inference, and defuzzification processes, the algorithm continuously adjusts the membership function and rule parameters to ultimately calculate the first experience optimization parameter set. The system feedback data acquisition module 104 is used to obtain multimodal feedback data generated by the system after the user performs operations in the system. The multimodal feedback data covers multiple dimensions such as visual feedback, auditory feedback, and operational feedback; The user experience detection module 105 is configured to input the multi-dimensional basic data and the multi-modal feedback data into a user experience precision detection model that has been pre-trained with massive user data and integrated with a reinforcement learning mechanism, and output the precise multi-dimensional experience value of the user in the current experience state; The experience optimization parameter adjustment module 106 is configured to use a multi-objective optimization algorithm to calculate a second set of experience optimization parameters based on the deviation between the precise multi-dimensional experience value and the preset multi-dimensional experience rating. This algorithm deeply adjusts the user experience to a second optimized state, thereby achieving continuous optimization of the user's full-cycle journey experience. The second optimized state represents an improvement in the user's overall experience throughout the full journey after the profound adjustments. By combining technologies such as multimodal feedback data processing and deep reinforcement learning, the particle swarm optimization (PSO) algorithm can be used to implement multi-objective optimization. The deviation between the precise multi-dimensional experience value and the preset multi-dimensional experience rating is used as multiple optimization targets, with particles representing different parameter combinations. Each particle adjusts its flight speed and position based on its own flight experience and the experience of the optimal particle in the swarm. Through continuous iteration, a parameter combination that optimizes multiple targets as much as possible is ultimately found, which is the second set of experience optimization parameters.

[0029] It should be noted that the user data acquisition module is used to obtain the user's multi-dimensional basic data, which includes the user's demographic information, psychological characteristic data, interactive behavior data and usage time. The multi-dimensional basic data mentioned here refers to a data set that describes user characteristics from different angles. These data can provide basic information for subsequent user demand prediction and experience optimization. Demographic information covers basic information such as the user's age, gender, and region. This information helps to conduct a preliminary classification and analysis of user groups; psychological characteristic data involves the user's emotional tendencies, cognitive style and other internal psychological states. By analyzing these data, we can better understand the user's behavioral motivations; interactive behavior data records the detailed process of the user's interaction with the system, such as clicks, slides and other operations, reflecting the user's usage habits and preferences; usage time directly reflects the user's attention and dependence on the system.

[0030] Specifically, the demographic information collection unit collects basic information such as the user's age, gender, and region through the detailed information filled in by the user during registration and subsequent supplementary information. For example, when registering, the user needs to fill in his or her date of birth, gender, and region, and this information will be stored in the database. The psychological characteristic data collection unit extracts psychological characteristic dimension data such as the user's emotional tendencies through the psychological assessment questionnaires, behavioral tasks, and interaction records with intelligent customer service that the user participates in in the system. For example, the answers to the emotional questionnaire completed by the user while using the system will be analyzed to determine the user's emotional state. The interactive behavior data collection unit collects multi-dimensional behavioral data such as the user's click behavior in real time through the system's built-in behavior monitoring module. This data will be used to construct a user behavior pattern map. For example, the system will record the number of clicks and dwell time of users in different functional modules to analyze the user's frequency of use and preference for each module.

[0031] Preferably, the user data acquisition module may also include a psychological profiling model unit. This unit inputs psychological characteristic data into a psychological profiling model based on a deep neural network. Using a long short-term memory (LSTM) network structure, it can capture the changing trends of a user's psychological characteristics over time and generate a psychological profile of the user. For example, by analyzing a user's long-term emotional data, the model can identify fluctuations in the user's emotions over different time periods, thereby providing a more accurate basis for personalized services.

[0032] Furthermore, the user churn prediction unit can use a Markov chain model to predict the probability of user churn and the likely churn time window based on the user's psychological profile and usage behavior patterns. For example, if a user frequently expresses anxiety and their usage frequency decreases over a period of time, the model will predict that the user may churn in the near future and take proactive measures to intervene.

[0033] In some embodiments, the user data acquisition module includes: Demographic information collection unit, used to collect the user's age, gender, region, educational background, occupation category and income level through the detailed information filled in by the user during registration and subsequent supplementary information; A user stratification model unit, configured to input the demographic information into a user stratification model based on cluster analysis, and divide the users into different segmented groups by combining the users' consumption behavior data and usage habit data; The demand preference prediction unit is used to use the Bayesian probability model to predict the user's demand preferences in the scenario based on the user's segmented group, the current usage scenario and time. The demand preferences include multiple dimensions such as demand preferences for functions, demand preferences for interface styles, and demand preferences for interaction methods.

[0034] It should be noted that the user data acquisition module includes a demographic information collection unit, which is used to collect the user's age, gender, region, educational background, occupation category, and income level through the detailed information filled in by the user during registration and subsequent supplementary information. The demographic information collection unit here refers to a system component specifically designed to collect basic user information, which is crucial for understanding the user's basic characteristics and needs. For example, age can help distinguish user groups of different age groups, gender can be used to analyze the impact of gender differences on product usage, and regional information helps understand user preferences in different regions.

[0035] Specifically, the demographic information collection unit can allow users to enter their age, gender and other information when registering in the form of an online form. For example, age can be classified according to age groups, such as 18-24 years old, 25-34 years old, etc., to facilitate subsequent analysis and processing. Gender can be set to options such as male, female, and other to meet the needs of different users. Regional information can be refined to different levels such as country, province, and city to more accurately analyze regional differences. Educational background can include academic level, such as high school, undergraduate, and graduate school. Occupational categories can cover different industries and positions, such as teachers, engineers, doctors, etc. Income levels can be set to different income ranges, such as less than 3,000 yuan, 3,000-5,000 yuan, etc.

[0036] Preferably, the user stratification model unit can use a cluster analysis algorithm to combine demographic information with the user's consumption behavior data and usage habit data to divide users into different segments. For example, the K-means clustering algorithm can be used to divide users into different groups such as high-value users, medium-value users, and low-value users based on their age, gender, consumption frequency, and other characteristics. The demand preference prediction unit can use a Bayesian probability model to predict the user's demand preferences in the scenario based on the segment group to which the user belongs and the current usage scenario and time. For example, when the user is in a shopping scenario, the demand preferences for product types, price ranges, etc. are predicted, thereby providing the user with more personalized services.

[0037] In some embodiments, the user data acquisition module further includes: The psychological characteristic data collection unit is used to extract psychological characteristic dimension data such as users' emotional tendencies, cognitive styles, motivation levels, and personality traits through natural language processing technology and sentiment analysis algorithms based on the psychological assessment questionnaires, behavioral tasks, and interaction records with intelligent customer service that users participate in in the system. A psychological profile model unit, configured to input the psychological characteristic data into a psychological profile model based on a deep neural network. The model adopts a long short-term memory network (LSTM) structure, which can capture the changing trend of the user's psychological characteristics over time and generate a psychological profile of the user; The user churn prediction unit is used to predict the probability of user churn and the possible churn time window using the Markov chain model based on the user's historical psychological state change trajectory and current usage behavior pattern if the user's psychological profile shows that he or she is anxious and has a low motivation level.

[0038] It should be noted that the user data acquisition module also includes a psychological characteristic data collection unit, which is used to extract psychological characteristic dimension data such as users' emotional tendencies, cognitive styles, motivation levels, and personality traits through psychological assessment questionnaires, behavioral tasks, and interaction records with intelligent customer service in the system using natural language processing technology and sentiment analysis algorithms. The psychological characteristic data collection unit mentioned here refers to a system component specifically designed to collect and analyze users' psychological states. This psychological characteristic data can help the system gain a deeper understanding of users' needs and behavioral motivations. For example, emotional tendencies can reflect the user's emotional state during use, cognitive style can reveal how users process information, motivation levels can indicate the user's inherent driving force for using the system, and personality traits can describe the user's personality traits.

[0039] Specifically, psychological assessment questionnaires can be designed to contain a series of questions that aim to assess the user's emotional state, cognitive preferences, and other psychological characteristics. For example, the questionnaire can include questions about the user's reaction under pressure, the user's acceptance of new things, and so on. Behavioral tasks can be specific activities designed in the system, and data is collected by observing the user's performance in these tasks. For example, a time management task can be used to assess the user's planning ability and self-control ability. Interaction records with intelligent customer service can be analyzed through natural language processing technology to analyze the user's language style and emotional expression. For example, analyze the vocabulary and tone used by users when communicating with customer service to determine the user's emotional tendencies.

[0040] Preferably, the psychological profiling model unit can utilize a deep neural network, specifically a long short-term memory (LSTM) network structure, to process and analyze psychological profile data. LSTM networks can effectively capture the changing trends of user psychological characteristics over time, thereby generating more accurate and dynamic user psychological profiles. For example, by analyzing user interaction records and behavioral task performance over different time periods, the LSTM model can identify patterns in user emotions and shifts in cognitive style.

[0041] Furthermore, the user churn prediction unit can combine psychological profiles with historical user behavior data, applying a Markov chain model to predict the probability of user churn. For example, if a user's psychological profile indicates chronic anxiety and declining motivation, while historical behavior data indicates a decrease in user login frequency, the model can predict the likelihood of user churn in the future and take proactive action to intervene.

[0042] In some embodiments, the user data acquisition module further includes: The interactive behavior data collection unit is used to collect multi-dimensional behavioral data such as user click behavior, sliding behavior, dwell time, operation path, and interaction frequency with system elements in various functional modules in real time through the system's built-in behavior monitoring module; A behavior pattern mining model unit is used to input the interactive behavior data into a behavior pattern mining model based on a graph neural network. The model can construct the user's behavior data into a behavior graph and mine the user's behavior pattern and preference path through graph convolution operations and node feature extraction; The function module optimization unit is used to optimize the operation process of a specific function module based on the business logic of the function module and the user's historical behavior data, using the Q-learning algorithm in reinforcement learning to improve the user's operation efficiency and experience satisfaction if the user's behavior pattern shows high frequency use of a specific function module and a relatively fixed path.

[0043] It should be noted that the user data acquisition module also includes an interactive behavior data collection unit, which is used to collect multi-dimensional behavioral data such as the user's click behavior, sliding behavior, dwell time, operation path, and interaction frequency with system elements in various functional modules in real time through the system's built-in behavior monitoring module. The interactive behavior data collection unit mentioned here refers to a system component specially designed to monitor and record user interaction behavior with the system. These behavioral data can reflect the user's usage habits and preferences. For example, click behavior can show the user's interest in a specific function, sliding behavior can reflect the user's way of browsing content, and dwell time can indicate the user's attention to certain pages or functions.

[0044] Specifically, the interactive behavior data collection unit can be implemented by embedding an event listener in the system. For example, when a user clicks a button, the event listener will record the time, location, and other information of the click behavior. Sliding behavior can be recorded by monitoring the user's gestures on the touch screen, including parameters such as the sliding direction and speed. The length of stay can be determined by calculating the time difference between the user entering and leaving a certain page or functional module. The operation path can record the user's navigation order in the system, for example, whether the user goes directly to the shopping cart from the homepage, or browses the product details page first. The frequency of interaction with system elements can count the number of times the user operates on a specific element, such as the frequency of the user clicking on a certain function button.

[0045] Preferably, the behavior pattern mining model unit can use a graph neural network (GNN) to process interactive behavior data. GNN can construct user behavior data into a behavior graph and, through graph convolution operations and node feature extraction, mine user behavior patterns and preferred paths. For example, by analyzing user transition paths between different functional modules, GNN can identify users' most frequently used operation processes.

[0046] Furthermore, the functional module optimization unit can use the Q-learning algorithm in reinforcement learning to optimize the operational process of functional modules based on user behavior patterns and historical behavior data. For example, if a user frequently performs repeated operations in a functional module, the Q-learning algorithm can adjust the interface layout or operation steps of the module to improve user operation efficiency.

[0047] In some embodiments, the user data acquisition module further includes: The usage time recording unit is used to accurately record the cumulative usage time, single usage time and the distribution of stay time in different functional modules after each user logs into the system through the system's time statistics module; A user value assessment model unit is used to combine the usage time data with the user's multi-dimensional basic data and interactive behavior data, and input the data into a user value assessment model based on a hybrid model. The model integrates a linear regression model and a nonlinear support vector machine model, comprehensively considering multiple dimensions such as the user's behavioral activity, loyalty, and consumption potential, to assess the user's comprehensive value level; The personalized service recommendation unit is used to provide users with customized service packages and recommended content based on the user's comprehensive value level, the current market environment and business goals, using dynamic pricing models and personalized recommendation algorithms to increase the user's usage time and consumption willingness.

[0048] It should be noted that the user data acquisition module also includes a usage time recording unit, which is used to accurately record the cumulative usage time, single usage time, and the distribution of stay time in different functional modules after each user logs into the system through the system's time statistics module. The usage time recording unit mentioned here refers to a component specially designed to monitor and record the time users use the system. These data can reflect the user's participation and interests in the system. For example, the cumulative usage time can show the user's overall investment in the system, the single usage time can reflect the user's concentration in a certain use, and the stay time distribution can reveal the user's preferences in different functional modules.

[0049] Specifically, the usage duration recording unit can be implemented by setting timestamps at the system login and exit points. For example, when a user logs into the system, the system records a start timestamp, and when the user exits the system, an end timestamp is recorded. The single usage duration is obtained by calculating the difference between the two timestamps. The cumulative usage duration can be obtained by adding up all single usage durations. The stay duration distribution can be obtained by recording timestamps when the user enters and leaves different functional modules, and calculating the difference. For example, a timestamp is recorded when the user enters the shopping cart module, and another is recorded when the user leaves. The difference between the two is the stay duration in the shopping cart module.

[0050] Preferably, the user value assessment model unit can use a hybrid model-based approach, combining a linear regression model and a nonlinear support vector machine model to assess the user's comprehensive value level. For example, a linear regression model can be used to assess the linear relationship between user behavioral activity and value, while a nonlinear support vector machine model can capture more complex nonlinear relationships, such as the interaction between user loyalty and consumption potential.

[0051] Furthermore, the personalized service recommendation unit can provide customized service packages and recommended content based on the user's comprehensive value level, combined with the market environment and business goals, using dynamic pricing models and personalized recommendation algorithms. For example, for high-value users, exclusive discount packages can be provided and high-value products can be prioritized; for low-value users, entry-level products and services can be recommended to increase their usage time and willingness to consume.

[0052] In some embodiments, the system feedback data acquisition module includes: The visual feedback data acquisition unit is used to capture the user's eye movement trajectory in real time through the embedded eye tracking device while the user is using the system, and record the user's gaze point, gaze duration, and scanning path of different interface elements; An interface visual perception model unit is used to combine the eye tracking data with the attribute information of the interface elements and input them into an interface visual perception model based on a convolutional neural network. The model adopts the Inception-ResNet architecture and learns the user's perception of the interface visual elements through multi-scale feature extraction and residual connections. The model outputs an interface color perception score, an element layout satisfaction score, and an icon recognition score. The auditory feedback data acquisition unit is used to collect the user's voice signal through the microphone array during the voice interaction between the user and the system, and record the sound effect signal output by the system; An auditory feedback analysis model unit is used to input the speech signal and sound effect signal into an auditory feedback analysis model based on deep speech recognition and audio feature extraction. The model uses a WaveNet architecture to generate high-quality speech features, and combines audio features such as Mel-frequency cepstral coefficients (MFCCs) and spectral centroids, using a support vector machine (SVM) classifier and regression model to evaluate sound effect quality, voice prompt clarity, and background music preference. The operation feedback data collection unit is used to collect the operation response time, number of erroneous operations and completion degree of the operation process in real time through the system performance monitoring module; The operation process analysis model unit is used to input the operation feedback data into the operation process analysis model based on operation process analysis. The model adopts the time series analysis method combined with the user experience quantitative model to evaluate the response speed, accuracy and convenience of the operation.

[0053] It should be noted that the system feedback data acquisition module includes a visual feedback data acquisition unit, which is used to capture the user's eye movement trajectory in real time through an embedded eye tracking device during the user's use of the system, and record the user's gaze point, gaze duration, and scan path of different interface elements. The visual feedback data acquisition unit mentioned here refers to a system component specially designed to collect user visual behavior data. These data can reflect the user's attention level and visual preferences for interface elements. For example, the gaze point can show the interface area that the user is most interested in, the gaze duration can reflect the depth of the user's attention to a specific element, and the scan path can reveal the order and pattern of the user's browsing of the interface.

[0054] Specifically, the visual feedback data acquisition unit can be implemented by embedding a high-precision eye-tracking device in the system. For example, the eye-tracking device can be installed on the user's display device to monitor the user's eye movements in real time. The gaze point can be recorded by the eye-tracking device where the user's eyes rest, and the gaze duration can be determined by calculating the time the user's eyes remain in a certain position. The scanning path can be depicted by connecting the movement trajectories of the user's eyes between different positions. The interface visual perception model unit can adopt a convolutional neural network (CNN)-based architecture, such as Inception-ResNet, to learn the user's perception patterns of the interface's visual elements through multi-scale feature extraction and residual connections. For example, the model can analyze the user's reaction to interface elements of different colors, shapes, and layouts, and output an interface color perception score, an element layout satisfaction score, and an icon recognition score.

[0055] Preferably, the auditory feedback data collection unit can utilize a microphone array to collect the user's voice signal and simultaneously record the system's sound effect output. For example, the microphone array can be placed around the user to capture high-quality voice data. The auditory feedback analysis model unit can utilize the WaveNet architecture to generate high-quality voice features. Combined with audio features such as Mel-Frequency Cepstral Coefficients (MFCCs) and spectral centroids, it employs a support vector machine (SVM) classifier and regression model to evaluate sound effect quality, voice prompt clarity, and background music preference. For example, the model can analyze user responses to different sound effects and voice prompts to optimize the system's auditory feedback. The operation feedback data collection unit can utilize the system performance monitoring module to collect real-time operation response time, number of incorrect operations, and completion of the operation process. For example, the system can record the response time for each user operation, count the number of incorrect operations, and evaluate the completion of the operation process. The operation process analysis model unit can utilize time series analysis methods, combined with a user experience quantification model, to evaluate the response speed, accuracy, and convenience of operations. For example, the model can analyze time series data of user operations to identify operational bottlenecks and areas for improvement.

[0056] In some embodiments, the output formula of the user demand dynamic prediction model is:

[0057] in, Represents the multi-dimensional prediction vector of user needs, is the weight matrix of the model, is the activation function, is the user's multi-dimensional basic data matrix, is the user's real-time interactive behavior data matrix, and are the bias vector and output bias vector of the model respectively; b is the bias vector of the model, and C is the bias vector of the model output; b is used to prevent overfitting, and c is used to adjust the final output of the model; The weight matrix and the bias vector 、 The adjustment formula is:

[0058]

[0059]

[0060] in, is the learning rate, represents the gradient operator, is the loss function, It is a multi-dimensional vector of users' actual needs.

[0061] Specifically, the activation function It can be a common function such as ReLU, sigmoid or tanh, which can introduce nonlinearity to the model and enable the model to learn complex mapping relationships. and the bias vector 、 The initial value of can be set randomly or initialized by some heuristic methods, such as Xavier initialization or He initialization, to accelerate the convergence of the model.

[0062] Furthermore, in practical applications, appropriate learning rates and loss functions can be selected through cross-validation and other methods to obtain the best model performance. In addition, in order to prevent the model from overfitting, regularization terms such as L1 or L2 regularization can be introduced to the weight matrix To constrain.

[0063] In some embodiments, the input formula of the user experience precision detection model is:

[0064] in, represents the input vector of the model, Provide users with multi-dimensional basic data. is the multimodal feedback data of the system, Real-time interactive behavior data of users in the system, As the data fusion function, the principal component analysis PCA and linear discriminant analysis LDA are combined to reduce the dimension and extract features of the input data.

[0065] Specific, multi-dimensional basic data Including user demographic information, psychological characteristics data, etc., which provide the basic characteristics and intrinsic needs of users. Multimodal feedback data of the system It covers multiple dimensions such as visual, auditory and operational feedback, reflecting the user's immediate experience of interacting with the system. Real-time interactive behavior data of users in the system The user's specific operations and behavior patterns are recorded. Data fusion function A combination of principal component analysis (PCA) and linear discriminant analysis (LDA) can be used to reduce the dimensionality of input data and extract features. For example, PCA can reduce the dimensionality of the data and remove redundant information, while LDA can enhance the class distinction of the data, enabling the model to learn and predict more effectively.

[0066] Preferably, the user experience precision detection model can use the Q-learning algorithm in deep reinforcement learning combined with the structure of a deep neural network. For example, a convolutional neural network (CNN) can be used to process image data, and a recurrent neural network (RNN) can be used to process sequence data to extract useful features. The prediction function of the model is According to the input feature vector and the model's parameter set To output the precise value of the user's multi-dimensional experience in the current experience state .For example, It can be a neural network containing multiple hidden layers, each hidden layer uses the ReLU activation function, and the output layer uses the softmax function to output the probability distribution of the experience value.

[0067] Furthermore, in order to improve the generalization ability of the model, dropout technology can be used to prevent overfitting. At the same time, the training process of the model can be optimized by adjusting hyperparameters such as learning rate and batch size.

[0068] In some embodiments, the output formula of the user experience precision detection model is:

[0069] in, Indicates the precise value of the user's multi-dimensional experience in the current experience state. is the prediction function of the model, is the parameter set of the model. The model adopts the Q-learning algorithm in deep reinforcement learning combined with the structure of deep neural network to continuously optimize the model parameters through interaction with users.

[0070] The multi-dimensional experience precision value mentioned here refers to the quantitative evaluation of the user's system experience from different perspectives, such as satisfaction and ease of use. The prediction function is the mathematical expression used in the model to calculate the output result based on the input data. The parameter set is the adjustable parameters in the model used to optimize the model's performance.

[0071] Specifically, the prediction function It can be designed as a deep neural network that contains multiple layers, such as input layer, hidden layer, and output layer. The number of neurons and the connection method of each layer can be set according to the specific problem. For example, the hidden layer can use the ReLU activation function to introduce nonlinearity, and the output layer can use the softmax function to output the probability distribution. Parameter set These parameters, including weights and biases, are adjusted by an optimization algorithm during the training process. Q-learning is a reinforcement learning method that learns optimal policies through interaction with the environment. In this model, Q-learning can be used to adjust the parameters of the neural network to maximize the user experience. For example, the model can use a reward mechanism to reinforce parameter adjustments that improve the user experience.

[0072] Preferably, to improve the accuracy and generalization ability of the model, some advanced techniques can be adopted. For example, batch normalization can be used to speed up the training process and reduce internal covariate shift.

[0073] Furthermore, regularization techniques, such as L2 regularization, can be introduced to prevent model overfitting. During training, cross-validation can be used to evaluate model performance and, based on the results, adjust hyperparameters such as the learning rate and hidden layer size. In addition, different optimization algorithms, such as Adam or RMSprop, can be considered, as these algorithms typically converge faster and perform better than traditional gradient descent.

[0074] The above-mentioned embodiments of the present invention have the following beneficial effects: the user full-cycle journey experience system described in the present invention can achieve accurate prediction and dynamic adjustment of user needs. Through deep learning models and multi-dimensional data input, the system can capture changes in user needs in real time, and calculate optimization parameters accordingly, and make preliminary adjustments to the user experience, so that users can obtain an experience that is more in line with their needs during use. In addition, the system can also obtain multimodal feedback data and use a precise detection model that integrates a reinforcement learning mechanism to accurately evaluate the multi-dimensional experience value of the user in the current experience state. Further, based on the deviation between the experience value and the preset rated value, a multi-objective optimization algorithm is used to calculate the deep optimization parameters, thereby achieving continuous optimization of the user's full-cycle journey experience and improving user satisfaction and loyalty.

[0075] The system also provides more comprehensive data support for demand forecasting and experience optimization through the multi-dimensional data collection of the User Data Acquisition Module, including demographic information, psychological characteristics, interactive behaviors, and usage duration. Furthermore, the System Feedback Data Acquisition Module collects multimodal data such as visual, auditory, and operational feedback. Combined with the deep learning algorithms of the User Experience Detection Module, this module can more accurately assess the user experience, enabling more effective adjustments to experience optimization parameters and ensuring continuous improvement of the user experience at every stage.

[0076] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement the above-mentioned embodiments, wherein the program instructions can be stored in the above-mentioned storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0077] The above descriptions merely illustrate some preferred embodiments of the present invention and the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A user full-cycle journey experience system, characterized by: include: User data acquisition module, used to obtain multi-dimensional basic data of users; A user demand prediction module is used to input the multi-dimensional basic data, the user's real-time interactive behavior data and the system usage frequency into a user demand dynamic prediction model based on deep learning, and output a multi-dimensional prediction vector of user demand; an experience optimization parameter calculation module, configured to calculate a first experience optimization parameter set using an adaptive adjustment algorithm based on the difference between the multi-dimensional prediction vector of the user demand and a preset user demand rating vector, so as to make preliminary adjustments to the user experience so that the user experience reaches a first optimized state; The system feedback data acquisition module is used to obtain multimodal feedback data generated by the system after the user operates in the system; A user experience detection module is used to input the multi-dimensional basic data and the multimodal feedback data into a user experience precision detection model that has been pre-trained with massive user data and integrated with a reinforcement learning mechanism, and output the precise multi-dimensional experience value of the user in the current experience state; The experience optimization parameter adjustment module is used to calculate a second experience optimization parameter set using a multi-objective optimization algorithm based on the deviation between the multi-dimensional experience precise value and the preset multi-dimensional experience rated value, so as to make the user experience reach the second optimized state and realize the continuous optimization of the user's full-cycle journey experience.

2. The user full-cycle journey experience system according to claim 1, characterized in that: The user data acquisition module includes: Demographic information collection unit, used to collect the user's age, gender, region, educational background, occupation category and income level through the detailed information filled in by the user during registration and subsequent supplementary information; A user stratification model unit, configured to input the demographic information into a user stratification model based on cluster analysis, and divide the users into different segmented groups by combining the users' consumption behavior data and usage habit data; The demand preference prediction unit is used to use the Bayesian probability model to predict the user's demand preferences in the scenario based on the user's segmented group, the current usage scenario and time. The demand preferences include multiple dimensions such as demand preferences for functions, demand preferences for interface styles, and demand preferences for interaction methods.

3. The user full-cycle journey experience system according to claim 2, characterized in that: The user data acquisition module also includes: The psychological characteristic data collection unit is used to extract psychological characteristic dimension data of users' emotional tendencies, cognitive styles, motivation levels, and personality traits through the psychological assessment questionnaires, behavioral tasks, and interaction records with intelligent customer service in the system using natural language processing technology and sentiment analysis algorithms; A psychological profile model unit, configured to input the psychological characteristic data into a psychological profile model based on a deep neural network; The user churn prediction unit is used to predict the probability of user churn and the churn time window using the Markov chain model based on the user's historical psychological state change trajectory and current usage behavior pattern if the user's psychological profile shows that he or she is anxious and has a low motivation level.

4. The user full-cycle journey experience system according to claim 3, characterized in that: The user data acquisition module also includes: The interactive behavior data collection unit is used to collect multi-dimensional behavioral data such as user click behavior, sliding behavior, dwell time, operation path, and interaction frequency with system elements in various functional modules in real time through the system's built-in behavior monitoring module; A behavior pattern mining model unit is used to input the interactive behavior data into a behavior pattern mining model based on a graph neural network, construct the user's behavior data into a behavior graph, and mine the user's behavior pattern and preference path through graph convolution operations and node feature extraction; The function module optimization unit is used to optimize the operation process of a specific function module using the Q-learning algorithm in reinforcement learning based on the business logic of the function module and the user's historical behavior data if the user's behavior pattern shows high frequency use of a specific function module and a relatively fixed path.

5. The user full-cycle journey experience system according to claim 4, characterized in that: The user data acquisition module also includes: The usage time recording unit is used to accurately record the cumulative usage time, single usage time and the distribution of stay time in different functional modules after each user logs into the system through the system's time statistics module; A user value evaluation model unit is used to combine the usage time data with the user's multi-dimensional basic data and interaction behavior data, and input them into a user value evaluation model based on a hybrid model; The personalized service recommendation unit is used to apply dynamic pricing models and personalized recommendation algorithms based on the user's comprehensive value level, the current market environment and business goals.

6. The user full-cycle journey experience system according to claim 5, characterized in that: The system feedback data acquisition module includes: The visual feedback data acquisition unit is used to capture the user's eye movement trajectory in real time through the embedded eye tracking device while the user is using the system, and record the user's gaze point, gaze duration, and scanning path of different interface elements; An interface visual perception model unit is used to combine the eye tracking data with the attribute information of the interface elements and input them into an interface visual perception model based on a convolutional neural network. The model adopts the Inception-ResNet architecture and learns the user's perception of the interface visual elements through multi-scale feature extraction and residual connections. The model outputs an interface color perception score, an element layout satisfaction score, and an icon recognition score. The auditory feedback data acquisition unit is used to collect the user's voice signal through the microphone array during the voice interaction between the user and the system, and record the sound effect signal output by the system; An auditory feedback analysis model unit is used to input the speech signal and sound effect signal into an auditory feedback analysis model based on deep speech recognition and audio feature extraction. The model uses a WaveNet architecture to generate high-quality speech features, and combines audio features such as Mel-frequency cepstral coefficients (MFCCs) and spectral centroids, using a support vector machine (SVM) classifier and regression model to evaluate sound effect quality, voice prompt clarity, and background music preference. The operation feedback data collection unit is used to collect the operation response time, number of erroneous operations and completion degree of the operation process in real time through the system performance monitoring module; The operation process analysis model unit is used to input the operation feedback data into the operation process analysis model based on operation process analysis. The model adopts the time series analysis method combined with the user experience quantitative model to evaluate the response speed, accuracy and convenience of the operation.

7. The user full-cycle journey experience system according to claim 6, characterized in that: The output formula of the user demand dynamic prediction model is: ； in, Represents the multi-dimensional prediction vector of user needs, is the weight matrix of the model, is the activation function, is the user's multi-dimensional basic data matrix, is the user's real-time interactive behavior data matrix, and are the bias vector and output bias vector of the model respectively; The weight matrix and the bias vector 、 The adjustment formula is: ； ； ； in, is the learning rate, represents the gradient operator, is the loss function, It is a multi-dimensional vector of users' actual needs.

8. The user full-cycle journey experience system according to claim 7, characterized in that: The input formula of the user experience precision detection model is: ； in, represents the input vector of the model, Provide users with multi-dimensional basic data. is the multimodal feedback data of the system, Real-time interactive behavior data of users in the system, As the data fusion function, the principal component analysis PCA and linear discriminant analysis LDA are combined to reduce the dimension and extract features of the input data.

9. The user full-cycle journey experience system according to claim 8, characterized in that: The output formula of the user experience precision detection model is: ； in, Indicates the precise value of the user's multi-dimensional experience in the current experience state. is the prediction function of the model, is the parameter set of the model. The model adopts the Q-learning algorithm in deep reinforcement learning combined with the structure of deep neural network to continuously optimize the model parameters through interaction with users.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the user full-cycle journey experience system as described in claims 1-9.

Images