A training method and system based on a display scheme data model

Abstract

The present invention discloses a training method and system for a display solution data model, including the following steps: S1, constructing a multi-source context dataset; S2, preprocessing the multi-source context dataset and integrating it into a single context vector; S3, generating a user preference profile and fusing the context vector with the user preference profile to generate a dynamic context preference profile; S4, inputting the dynamic context preference profile into a conditional generative adversarial network to generate a display solution; S5, embedding an adaptive neural regulation module to optimize the display solution generation effect; S6, introducing a hierarchical context control mechanism; S7, iteratively updating the user preference profile based on historical interaction feedback; S8, analyzing the similarity between the current context and the historical context, applying a meta-learning algorithm to determine the applicability of the display solution, and migrating the effective display solutions in the historical context to the current context. The present invention adopts a model training method of multi-dimensional context awareness and dynamic feedback optimization, realizing the context-adaptive generation of display solutions.

Description

Technical Field

[0001] The present invention relates to the technical field of data model training, and particularly to a method and system for training a data model based on a display scheme. Background Art

[0002] In the current display technology field, personalized recommendation and intelligent display have become important means to enhance the user experience. Traditional display systems mainly rely on static rules or display algorithms based on simple user preferences, usually using basic user information (such as age, gender, browsing history, etc.) for simple content recommendation and display layout. However, this static and relatively single display method has many deficiencies in the user experience, especially in multi - context and high - interactivity scenarios, it is difficult to fully meet the diverse needs of users. With the development of intelligent devices and Internet of Things technology, user interaction behavior data, device information, and environmental parameters can be collected in real - time, which provides a basis for the further optimization of intelligent display systems. However, there are still many deficiencies in the existing technology in making full use of multi - dimensional context data, dynamic feedback, and multi - level control, etc.

[0003] The deficiencies of traditional display systems are mainly reflected in context adaptability and personalized needs. Display systems usually have difficulty in real - time perceiving and adapting to changes in the user context. Existing display systems usually only consider the user's historical data and ignore the user's immediate needs in different contexts. For example, the preferences of the same user may vary between day and night, indoor and outdoor, but existing display systems are difficult to flexibly capture and adapt to these changes. In addition, the user's interest preferences may change over time, while the recommendation algorithms of traditional systems often lack the mechanism to identify and adaptively update to the fluctuations of user interests, and it is difficult to adjust the recommended content in a timely manner. Therefore, after long - term use of the same system by users, the matching degree of the display content may gradually decrease, resulting in a reduction in user experience and user satisfaction.

[0004] In the field of intelligent display, the application of generative adversarial networks (GANs) and deep learning technologies has been gradually introduced to generate diverse display content. However, the existing application of GANs also has limitations, mainly lacking dynamic adaptability to the context. Most GAN models rely on fixed training data sets and model parameters when generating content, and it is difficult to adjust the characteristics of the generated content in real - time according to the user's current context. In actual scenarios, user interaction behaviors (such as clicks, dwell time, sliding behaviors, etc.) can provide rich feedback information, and this feedback information can guide the generator to adjust the content generation strategy to better meet the user's immediate needs. However, the existing GAN generation technology has not fully utilized the user's interaction feedback to dynamically update the generation model, resulting in deficiencies in the real - time matching of the generated content with the user context.

[0005] In addition, traditional display systems usually adopt a single-layer context control method, making it difficult to achieve hierarchical control in complex scenarios. Users' needs often have multi-level characteristics, and the requirements for the theme, style, and details of the displayed content may vary in different scenarios. For example, users need concise and clear information display in the news scenario, while they may prefer visual product displays in the shopping scenario; users have higher requirements for brightness in the daytime environment, while they may prefer a dark theme when browsing at night. Existing display systems usually can only perform single control based on users' main needs and cannot perform layer-by-layer matching at multiple levels such as display theme, style, and details, resulting in insufficient adaptability of the displayed content in specific scenarios and the failure to maximize the user experience and interaction effects.

[0006] In terms of feedback optimization, existing technologies usually lack self-optimization mechanisms and it is difficult to iteratively update the display content generation process based on users' real-time behavior feedback. Most display systems lack perfect feedback loops and adaptive optimization mechanisms, usually only providing display content passively and unable to optimize recommendation strategies based on users' behavior feedback. For example, when the click-through rate or dwell time of users decreases, the system cannot adjust the display content generation strategy in real time, so it cannot improve the relevance of the content and the user experience in a timely manner. In addition, existing optimization mechanisms are mainly based on preset rules or simple statistical data and it is difficult to perform in-depth optimization according to users' complex behavior patterns. Therefore, the improvement of the user experience is usually intermittent and it is difficult to form continuous adaptive optimization.

[0007] In terms of cross-scenario migration and expansion, the scalability of existing display systems is relatively limited and it is difficult to quickly adapt to new requirements in different scenarios. Each change in the scenario may require the system to retrain and adjust the display model again, consuming a large amount of time and resources, resulting in the difficulty of quickly adapting the display solution in dynamic scenarios. The introduction of meta-learning and transfer learning technologies can improve the adaptability of the model to a certain extent, but the application of these technologies in existing systems is relatively preliminary and it is impossible to achieve rapid migration and expansion of the display solution. For example, in similar scenarios, the current display system cannot quickly migrate historical display strategies to new scenarios, resulting in low adaptability of the system in new scenarios. This not only increases the training overhead of the system in multiple scenarios but also limits the rapid response ability of the display system under multiple scenarios and multiple requirements.

[0008] Therefore, how to provide a training method and system based on a display solution data model is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0009] An object of the present invention is to provide a training method and system for a display scheme data model. Through multi-dimensional context awareness, adaptive generation, and feedback optimization methods, the present invention achieves a high degree of matching between display content and user context. By combining hierarchical context control and conditional generative adversarial networks, the generated content is adjusted in real time according to the user's click-through rate, dwell time, and sliding behavior, achieving personalized and adaptive display effects. At the same time, by combining meta-learning and transfer learning technologies, the system can quickly respond to user needs in multiple contexts, with the advantages of high matching degree, fast response speed, and strong scalability, providing an innovative solution for intelligent display.

[0010] A training method for a display scheme data model according to an embodiment of the present invention includes the following steps:

[0011] S1. Collect the user's micro-context data through Internet of Things devices, extract the macro-context data from the user device information, and construct a multi-source context data set;

[0012] S2. Preprocess the multi-source context data set, weight and fuse the context features from different sources through a multi-head attention mechanism, and integrate the multi-source context features into a single context vector;

[0013] S3. Extract the preference features from the user's historical interaction data, generate a user preference profile, and fuse the context vector with the user preference profile to generate a dynamic context preference profile;

[0014] S4. Use the generated dynamic context preference profile as a condition to input into a conditional generative adversarial network to generate a display scheme that conforms to the current context;

[0015] S5. Embed an adaptive neural regulation module, use the context feature feedback loop to dynamically adjust the generation strategy of the conditional generative adversarial network, and adjust the weights of the generation model in real time according to the user's click-through rate, dwell time, and sliding behavior to optimize the display scheme generation effect;

[0016] S6. Introduce a hierarchical context control mechanism, and the hierarchical context control mechanism includes macro-context control, meso-context control, and micro-context control;

[0017] S7. Collect the feedback data of user interaction, input the feedback data into the dynamic context preference profile and the adaptive neural regulation module, adjust the weight parameters of the conditional generative adversarial network, and iteratively update the user preference profile based on the historical interaction feedback;

[0018] S8. Analyze the similarity between the current context and the historical context through context similarity detection, apply a meta-learning algorithm to determine the applicability of the display scheme, and use transfer learning to transfer the effective display scheme in the historical context to the current context.

[0019] Optionally, S4 specifically includes:

[0020] S41. Obtain a dynamic context preference portrait , where represents the th feature dimension value in the dynamic context preference portrait;

[0021] S42. Input the dynamic context preference portrait into the generator of the conditional generative adversarial network. The generator maps to a hidden layer vector , and combines it with a generated noise vector to generate a preliminary display content vector :

[0022] ;

[0023] Wherein, represents the hyperbolic tangent activation function, represents the adaptive adjustment parameter, represents the weight matrix of the dynamic context preference portrait, represents the weight matrix of the hidden layer vector, represents the weight matrix of the noise vector, represents the bias term, represents the number of feature dimensions of the dynamic context preference portrait, represents the dimension number of the hidden layer vector, represents the dimension number of the noise vector, represents the smoothing term, , and respectively represent each weight coefficient in the weight matrices of the dynamic context preference portrait, the hidden layer vector, and the noise vector, represents the th feature dimension value in the hidden layer vector, represents the th feature dimension value in the noise vector;

[0024] S43. According to the processing of the preliminary display content vector , output a preliminary display scheme. The preliminary display scheme includes layout elements, color matching elements, and content elements. The layout elements include button positions and image display area sizes. The color matching elements include background colors and text colors. The content elements include the type, style, and text content of the displayed content;

[0025] S44. Use the discriminator to evaluate the output of the generator. The input conditions are the dynamic context preference portrait and the preliminary display content vector , the discriminator determines the generated solution based on the matching degree of the displayed content and outputs a matching degree score:

[0026] ;

[0027] Among them, represents the matching degree score, represents the exponential function, represents the weight matrix in the discriminator, represents the context vector weight in the discriminator, represents the bias term, represents the dynamic context preference portrait and the combined feature vector of the preliminary display content vector ; represents the feature vector of the dynamic context preference portrait ; represents each weight term in the context vector weight, represents the preliminary display content vector in each weight term, represents the preliminary display content vector in each eigenvalue, represents the number of feature dimensions of the context vector, represents the preliminary display content vector ;

[0028] S45. If the matching degree score does not reach the set matching threshold, optimize the weight matrix and bias value of the generator through backpropagation to generate a display content that better conforms to the current context;

[0029] S46. Through multiple iterations, during the alternating training process of the generator and the discriminator, the display solution generated by the generator gradually conforms to the dynamic context preference portrait , and finally output a display solution that conforms to the current context.

[0030] Optionally, the S5 specifically includes:

[0031] S51. Collect the real-time interaction data of the user, including the click-through rate , the dwell time and the sliding behavior , the click-through rate represents the user's interest in the displayed content, the dwell time reflects the user's attention, and the sliding behavior represents the user's tendency to explore the displayed content and the interaction area preference;

[0032] S52. For the click-through rate , Dwell time and sliding behavior are standardized and converted into a unified feedback score:

[0033] ;

[0034] Among them, represents the unified feedback score, represents the weight coefficient of the click-through rate, represents the weight coefficient of the dwell time, represents the weight coefficient of the sliding behavior, represents the smoothing coefficient, represents the weight parameter of the weight of the layer in the conditional generative adversarial network, represents the eigenvalue of the layer in the conditional generative adversarial network;

[0035] S53. Input the unified feedback score into the adaptive neural regulation module, and use the unified feedback score to adjust the weight matrix of the generator of the conditional generative adversarial network;

[0036] S54. Use the gradient descent method to update the weight matrix of the generator, the hidden layer vector weight matrix, the noise vector weight matrix, and the bias value in the conditional generative adversarial network in real time;

[0037] S55. Introduce the regulation parameters , and of the adaptive feedback loop, where represents the response sensitivity of the click-through rate, represents the response sensitivity of the dwell time, represents the response sensitivity of the sliding behavior;

[0038] S56. Through the weight update and iteration of the feedback loop, optimize the generation effect of the display scheme. When the click-through rate, dwell time, and sliding behavior reach the predetermined threshold, it is determined that the generated display scheme meets the current scenario requirements.

[0039] Optionally, the S6 specifically includes:

[0040] S61. Identify the user's current main demand scenario, determine the theme and overall direction of the display content according to the user behavior data and preference profile. The macro scenario control module analyzes the user's scenario requirements and adjusts the theme of the display content through the scenario weight so that the display content matches the user's demand scenario;

[0041] S62. Adapt and adjust the content style by combining user device information and environmental factors. The mesoscopic context control module applies corresponding style templates based on different device and environmental characteristics, and uses the mesoscopic context style weight coefficient to control style conversion;

[0042] S63. When the user interacts with the displayed content, real-time detect the user's click, swipe, and stay behaviors, and adjust the details of the displayed content according to the microscopic interaction behaviors. The microscopic context control module dynamically updates the displayed content according to the user's real-time feedback, including adjusting button colors, content layout and formatting, through the microscopic context control weight coefficient to control the adjustment amplitude of the display details;

[0043] S64. According to the hierarchical context control mechanism, define the multi-level fitness score of the displayed content to measure the adaptability of the displayed content at different context levels:

[0044] ;

[0045] where, represents the multi-level fitness score of the displayed content, represents the macro context control weight coefficient, represents the macro context fitness, represents the micro context fitness, represents the weight parameter of the th feature in the macro context control, represents the th feature in the macro context control, represents the number of features in the macro context control, represents the weight parameter of the th feature in the mesoscopic context control, represents the th feature in the mesoscopic context control, represents the number of features in the mesoscopic context control, represents the weight parameter of the th feature in the microscopic context control, represents the th feature in the microscopic context control, represents the number of features in the microscopic context control;

[0046] S65. When the fitness score reaches the preset threshold, it is considered that the displayed content has achieved the expected context matching effect at the macro, mesoscopic, and microscopic context levels, and it is determined that the displayed content meets the user's multi-level context requirements;

[0047] S66. If the fitness score If the preset threshold is not reached, the weight coefficients of the macro - situation control module, the meso - situation control module, and the micro - situation control module are updated through the back - propagation mechanism, and the adaptability of the displayed content is gradually optimized.

[0048] Optionally, S7 specifically includes:

[0049] S71. Collect comprehensive interaction data of the user in the displayed content, including click - through rate, browsing duration, sliding frequency, and hot - spot areas, to form multi - dimensional interaction data;

[0050] S72. Convert the collected multi - dimensional interaction data into a situation feedback vector, and after data cleaning and pre - processing, fuse it with the user's historical portrait data. By analyzing the preference fluctuations of the user in different situations, generate a personalized vector with time and situation specificity;

[0051] S73. The dynamic situation preference portrait module generates a multi - stage user portrait update strategy according to the input of the situation feedback vector. The update strategy separates the short - term preference from the long - term preference based on the user's interaction frequency, time interval, and changes in situation requirements, and constructs a preference feature vector that conforms to the user's recent interests;

[0052] S74. The adaptive neural regulation module receives the updated situation preference portrait information and automatically detects the deviation area of the user's interest characteristics;

[0053] S75. Through the multi - level feedback control mechanism of the adaptive neural regulation module, further introduce a preference adjustment weight, so that the generator automatically adjusts the display scheme of the displayed content. The preference adjustment weight is based on the user's situation feedback frequency and situation change trend, and dynamically regulates the style, color matching, and structure display characteristics of the generated content;

[0054] S76. In multiple rounds of iterative optimization, by recording the generation effect after each user interaction, gradually construct a historical preference map in the user portrait, accumulate the user's preference change trajectory, and based on the historical preference map generator, identify the long - term pattern of the user's preferences, and give priority to referring to the long - term pattern of the user's preferences when generating content.

[0055] Optionally, S8 specifically includes:

[0056] S81. Collect the current user situation data, including the user's real - time behavior characteristics, device information, and environmental parameters, to form the current situation feature vector ;

[0057] S82. Based on the situation similarity detection model, calculate the current situation feature vector and the historical situation feature vector set The similarity between them is calculated based on cosine similarity or Euclidean distance metric method, and the feature vector most similar to the current situation is selected;

[0058] S83. Apply the meta - learning algorithm to determine the adaptability of the display scheme in the historical situation to the current situation:

[0059] ;

[0060] Among them, represents the applicability score, represents the number of historical situation features, represents the th weight of the feature, represents the th weight of the feature, represents the th feature value of the current situation, represents the th feature value of the historical situation, represents the exponential function, represents the exponential decay coefficient, represents the smoothing coefficient, represents the non - linear adjustment coefficient, represents the th feature value of the current situation, represents the th feature value of the historical situation;

[0061] S84. For the display scheme determined to be applicable, the transfer learning module transfers the generation model parameters in the historical situation to the current situation. The transfer process includes the selection of key parameters, the conversion of feature mapping, and the optimization of content features;

[0062] S85. For the transferred display scheme, the generation model is fine - tuned. The fine - tuning process adjusts the display strategy, color - matching style, and interaction details in the model based on the differences between the current situation and the historical situation;

[0063] S86. After completing the transfer and fine - tuning, record and save the generation effect in the current situation and the user feedback data as new historical situation features.

[0064] A display scheme data model training system according to an embodiment of the present invention includes the following modules:

[0065] The data acquisition module is used to collect the microscopic situation data of users through Internet of Things devices, extract the macroscopic situation data from the user device information, and generate a situation data set containing multi - source situation features;

[0066] A data preprocessing module for cleaning, standardizing, and denoising the scenario dataset, and generating a single scenario vector by weighted fusion of scenario features from different sources through a multi-head attention mechanism;

[0067] A user profile generation module for extracting preference features from the user's historical interaction data, generating a user preference profile, and fusing the scenario vector with the user preference profile to generate a dynamic scenario preference profile;

[0068] A conditional generative adversarial network module, including a generator and a discriminator, for receiving the dynamic scenario preference profile as a conditional input and generating a display scheme that conforms to the current scenario;

[0069] An adaptive neural regulation module for dynamically adjusting the generation strategy of the conditional generative adversarial network according to the real-time feedback of the user's click-through rate, dwell time, and swiping behavior, and optimizing the generation effect of the display scheme through a feedback loop;

[0070] A hierarchical scenario control module, including a macro scenario control module, a meso scenario control module, and a micro scenario control module, for adjusting the display scheme layer by layer according to user needs, device information, and real-time interaction behavior;

[0071] A feedback data processing module for collecting and processing the user's interaction feedback data, inputting the feedback data into the dynamic scenario preference profile and the adaptive neural regulation module, and iteratively updating the user preference profile based on historical interaction feedback;

[0072] A scenario similarity detection module for analyzing the similarity between the current scenario and historical scenarios based on a scenario similarity detection model, determining the applicability of historical display schemes in combination with a meta-learning algorithm, and migrating highly applicable historical display schemes to the current scenario through a transfer learning module;

[0073] A storage module for recording and saving the generation effect and user feedback data in the current scenario after the migration and fine-tuning of the display scheme, and updating the scenario feature database.

[0074] The beneficial effects of the present invention are as follows:

[0075] First, the present invention can collect the user's micro scenario data and macro scenario data in real time, integrate multi-source scenario features into a single scenario vector through a multi-head attention mechanism, and generate a dynamic scenario preference profile based on the user's historical interaction data, so as to more accurately understand the user's preferences and the needs of the current scenario. Through the conditional generative adversarial network, the system can dynamically consider the user scenario when generating display content, making the content highly match the user's needs, which effectively makes up for the deficiency of the traditional display system in terms of scenario adaptability.

[0076] Secondly, by embedding an adaptive neural regulation module and introducing a feedback loop mechanism into the conditional generative adversarial network, the present invention can dynamically adjust the weights of the generative model based on real-time feedback such as the user's click-through rate, dwell time, and swiping behavior, thereby realizing the self-optimization of the display scheme. The introduction of this feedback mechanism enables the system to quickly respond to changes in user behavior and optimize the generation strategy of the display content according to the interaction feedback, avoiding the drawbacks of relying on fixed rules in traditional systems, achieving continuous adaptive updates, and enhancing the overall user experience.

[0077] In addition, the application of the hierarchical context control mechanism enables the system to perform content control at three levels: macro, meso, and micro. It gradually adjusts the theme, style, and details of the display content to meet the diverse needs of users at different context levels. Different from traditional display systems that usually only focus on single-level context control, the hierarchical context control mechanism of the present invention greatly improves the adaptability of the display content through multi-level regulation at the macro, meso, and micro levels, enabling users to obtain a display scheme that better conforms to their own preferences in different demand contexts. This multi-level context control breaks through the limitations of the prior art and brings a smoother and more natural interaction experience to users.

[0078] In terms of cross-context transfer, the present invention uses a context similarity detection model and a meta-learning algorithm to quickly identify the similarity between the current context and historical contexts, and transfers applicable historical display schemes to the new context to achieve the rapid generation and adaptation of display content. The application of transfer learning not only reduces the need for re-training the display model in the new context but also significantly improves the system's response speed and scalability in multiple contexts. Regardless of how the user's context changes, the system can quickly adapt to the new display scenario by transferring effective historical solutions, overcoming the deficiencies of traditional display systems in multi-context adaptability. This flexible adaptability enables the system to always maintain a high degree of matching of the display content, while significantly reducing resource consumption and enhancing the intelligence and efficiency of the display system. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0080] Figure 1 is the overall flowchart of a training method for a display scheme data model proposed by the present invention;

[0081] Figure 2 is the structural schematic diagram of a training system for a display scheme data model proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0082] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, and thus only showing the components related to the present invention.

[0083] Reference Figure 1 , a training method for a display scheme data model, comprising the following steps:

[0084] S1. Collect microscopic context data of users through Internet of Things devices, extract macroscopic context data from user device information, and construct a multi-source context dataset;

[0085] S2. Preprocess the multi-source context dataset, weight and fuse context features from different sources through a multi-head attention mechanism, and integrate the multi-source context features into a single context vector;

[0086] S3. Extract preference features from the historical interaction data of users, generate a user preference portrait, and fuse the context vector with the user preference portrait to generate a dynamic context preference portrait;

[0087] S4. Use the generated dynamic context preference portrait as a condition to input into a conditional generative adversarial network to generate a display scheme that conforms to the current context;

[0088] S5. Embed an adaptive neural regulation module, use the context feature feedback loop to dynamically adjust the generation strategy of the conditional generative adversarial network, and adjust the weights of the generation model in real time according to the user's click-through rate, dwell time, and sliding behavior to optimize the generation effect of the display scheme;

[0089] S6. Introduce a hierarchical context control mechanism, and the hierarchical context control mechanism includes macroscopic context control, mesoscopic context control, and microscopic context control;

[0090] S7. Collect feedback data of user interactions, input the feedback data into the dynamic context preference portrait and the adaptive neural regulation module, adjust the weight parameters of the conditional generative adversarial network, and iteratively update the user preference portrait based on historical interaction feedback;

[0091] S8. Analyze the similarity between the current context and the historical context through context similarity detection, apply a meta-learning algorithm to determine the applicability of the display scheme, and use transfer learning to transfer the effective display scheme in the historical context to the current context.

[0092] In this embodiment, S4 specifically includes:

[0093] S41. Obtain the dynamic context preference portrait , where represents the value of the th feature dimension in the dynamic context preference portrait;

[0094] S42. Input the dynamic context preference portrait into the generator of the conditional generative adversarial network. The generator maps it to a hidden layer vector and combines it with a generated noise vector to generate a preliminary display content vector :

[0095] ;

[0096] Among them, represents the hyperbolic tangent activation function, represents the adaptive adjustment parameter, represents the weight matrix of the dynamic context preference portrait, represents the weight matrix of the hidden layer vector, represents the weight matrix of the noise vector, represents the bias term, represents the number of feature dimensions of the dynamic context preference portrait, represents the dimension number of the hidden layer vector, represents the dimension number of the noise vector, represents the smoothing term, , and respectively represent each weight coefficient in the weight matrices of the dynamic context preference portrait, hidden layer vector, and noise vector, represents the value of the th feature dimension in the hidden layer vector, represents the value of the th feature dimension in the noise vector;

[0097] S43. According to the processing of the preliminary display content vector , output a preliminary display plan. The preliminary display plan includes layout elements, color matching elements, and content elements. Among them, the layout elements include button positions and the sizes of image display areas, the color matching elements include background colors and text colors, and the content elements include the types, styles, and text contents of the display content;

[0098] S44. Use the discriminator to evaluate the output of the generator. The input conditions are the dynamic context preference portrait and the preliminary display content vector . The discriminator determines the generation plan based on the matching degree of the display content and outputs a matching degree score:

[0099] ;

[0100] Among them, represents the matching degree score, represents the exponential function, Denotes the weight matrix in the discriminator, Denotes the context vector weights in the discriminator, Denotes the bias term, Denotes the dynamic context preference portrait And the combined feature vector of the preliminary display content vector ; Denotes the feature vector of the dynamic context preference portrait ; Denotes each weight term in the context vector weights, Denotes the preliminary display content vector Each weight term in; Denotes the preliminary display content vector Each eigenvalue in; Denotes the number of feature dimensions of the context vector, Denotes the preliminary display content vector The number of feature dimensions of;

[0101] S45. If the matching degree score Does not reach the set matching threshold, perform forward propagation optimization on the weight matrix and bias value of the generator to generate a display content that better conforms to the current context;

[0102] S46. Through multiple iterations, the display scheme generated by the generator gradually conforms to the dynamic context preference portrait During the alternating training process of the generator and the discriminator, and finally output a display scheme that conforms to the current context.

[0103] In this embodiment, the S5 specifically includes:

[0104] S51. Collect the real-time interaction data of the user, including the click-through rate , the dwell time And the sliding behavior , the click-through rate Represents the user's interest in the display content, the dwell time Reflects the user's attention, and the sliding behavior Represents the tendency of the user to explore the display content and the interaction area preference;

[0105] S52. Standardize the click-through rate , the dwell time And the sliding behavior And convert them into a unified feedback score:

[0106] ;

[0107] Among them, Represents the unified feedback score, Represents the weight coefficient of the click-through rate, Represents the weight coefficient of the dwell time, Represents the weight coefficient of the sliding behavior, Represents the smoothing coefficient, Represents the weight parameter of the weight of the layer in the conditional generative adversarial network, Represents the eigenvalue of the layer in the conditional generative adversarial network;

[0108] S53. Input the unified feedback score into the adaptive neural regulation module, and use the unified feedback score to adjust the weight matrix of the generator of the conditional generative adversarial network;

[0109] S54. Use the gradient descent method to update the weight matrix of the generator, the weight matrix of the hidden layer vector, the weight matrix of the noise vector, and the bias value in the conditional generative adversarial network in real time;

[0110] S55. Introduce the regulation parameters , and in the adaptive feedback loop, where represents the response sensitivity of the click-through rate, represents the response sensitivity of the dwell time, represents the response sensitivity of the sliding behavior;

[0111] S56. Through the weight update and iteration of the feedback loop, optimize the generation effect of the display scheme. When the click-through rate, dwell time, and sliding behavior reach the predetermined thresholds, it is determined that the generated display scheme meets the current scenario requirements.

[0112] In this embodiment, the specific steps of S6 are as follows:

[0113] S61. Identify the user's current main demand scenario, determine the theme and overall direction of the display content according to the user behavior data and preference profile. The macro scenario control module analyzes the user's scenario requirements and adjusts the theme of the display content through the scenario weight to make the display content match the user's demand scenario;

[0114] S62. Adaptively adjust the content style in combination with the user device information and environmental factors. The meso scenario control module applies the corresponding style template based on different device and environmental characteristics and uses the meso scenario style weight coefficient to control the style conversion;

[0115] S63. When the user interacts with the displayed content, the click, swipe, and stay behaviors of the user are detected in real time, and the details of the displayed content are adjusted according to the micro-interaction behaviors. The micro-context control module dynamically updates the displayed content according to the real-time feedback of the user, including adjusting the button color, content layout, and layout, and controlling the adjustment range of the display details through the micro-context control weight coefficient Control the adjustment range of the display details;

[0116] S64. According to the hierarchical context control mechanism, define the multi-level fitness scores of the displayed content, which are used to measure the adaptability of the displayed content at different context levels:

[0117] ;

[0118] Among them, Represents the multi-level fitness score of the displayed content, Represents the macro-context control weight coefficient, Represents the macro-context fitness, Represents the micro-context fitness, Represents the weight parameter of the th feature in the macro-context control, Represents the th feature in the macro-context control, Represents the number of features in the macro-context control, Represents the weight parameter of the th feature in the meso-context control, Represents the th feature in the meso-context control, Represents the number of features in the meso-context control, Represents the weight parameter of the th feature in the micro-context control, Represents the th feature in the micro-context control, Represents the number of features in the micro-context control;

[0119] S65. When the fitness score reaches the preset threshold, it is considered that the displayed content has achieved the expected context matching effect at the macro, meso, and micro context levels, and it is determined that the displayed content meets the multi-level context needs of the user;

[0120] S66. If the fitness score does not reach the preset threshold, the weight coefficients of the macro-context control module, meso-context control module, and micro-context control module are updated through the backpropagation mechanism to gradually optimize the adaptability of the displayed content.

[0121] In this embodiment, the specific content of S7 includes:

[0122] S71. Collect comprehensive interaction data of users in the displayed content, including click-through rate, browsing duration, sliding frequency, and hot areas, to form multi-dimensional interaction data;

[0123] S72. Convert the collected multi-dimensional interaction data into a context feedback vector, and after data cleaning and preprocessing, fuse it with the user's historical portrait data. By analyzing the preference fluctuations of users in different contexts, generate a personalized vector with time and context specificity;

[0124] S73. The dynamic context preference portrait module generates a multi-stage user portrait update strategy based on the input of the context feedback vector. The update strategy separates the short-term preference from the long-term preference according to the user's interaction frequency, time interval, and the change of context requirements, and constructs a preference feature vector that conforms to the user's recent interests;

[0125] S74. The adaptive neural regulation module receives the updated context preference portrait information and automatically detects the deviation area of the user's interest characteristics;

[0126] S75. Through the multi-level feedback control mechanism of the adaptive neural regulation module, further introduce a preference adjustment weight, so that the generator automatically adjusts the display scheme of the displayed content. The preference adjustment weight is based on the user's context feedback frequency and context change trend, and dynamically regulates the style, color matching, and structure display characteristics of the generated content;

[0127] S76. In multiple rounds of iterative optimization, by recording the generation effect after each user interaction, gradually construct a historical preference map in the user portrait, accumulate the preference change trajectory of the user, and based on the historical preference map generator, identify the long-term pattern of the user's preferences, and give priority to referring to the long-term pattern of the user's preferences when generating content.

[0128] In this embodiment, the specific steps of S8 include:

[0129] S81. Collect the current user context data, including the user's real-time behavior characteristics, device information, and environmental parameters, to form the current context feature vector ;

[0130] S82. Based on the context similarity detection model, calculate the similarity between the current context feature vector and the set of historical context feature vectors . The similarity calculation is based on the cosine similarity or Euclidean distance metric method, and filter out the feature vector that is most similar to the current context;

[0131] S83. Apply the meta-learning algorithm to determine the adaptability of the display scheme in the historical context to the current context:

[0132] ;

[0133] wherein, represents the applicability score, represents the number of historical context features, represents the th feature weight, represents the th feature weight, represents the th feature value of the current context, represents the th feature value of the historical context, represents the exponential function, represents the exponential decay coefficient, represents the smoothing coefficient, represents the non - linear adjustment coefficient, represents the th feature value of the current context, represents the th feature value of the historical context;

[0134] S84. For the display scheme determined to be applicable, the transfer learning module transfers the generation model parameters in the historical context to the current context. The transfer process includes the selection of key parameters, the conversion of feature mapping, and the optimization of content features;

[0135] S85. For the transferred display scheme, the generation model is fine - tuned. The fine - tuning process adjusts the display strategy, color - matching style, and interaction details in the model based on the differences between the current context and the historical context;

[0136] S86. After completing the transfer and fine - tuning, record and save the generation effect in the current context and the user feedback data as new historical context features.

[0137] Reference Figure 2 , a display scheme data model - based training system, includes the following modules:

[0138] The data acquisition module is used to collect the user's micro - context data through Internet of Things devices, extract the macro - context data from the user device information, and generate a context data set containing multi - source context features;

[0139] The data pre - processing module is used to perform data cleaning, standardization, and denoising processing on the context data set, and perform weighted fusion on context features from different sources through the multi - head attention mechanism to generate a single context vector;

[0140] The user portrait generation module is used to extract preference features from the user's historical interaction data, generate a user preference portrait, and fuse the context vector with the user preference portrait to generate a dynamic context preference portrait;

[0141] The conditional generative adversarial network module, including a generator and a discriminator, is used to receive the dynamic context preference portrait as a conditional input and generate a display scheme that conforms to the current context;

[0142] The adaptive neural regulation module is used to dynamically adjust the generation strategy of the conditional generative adversarial network according to the real-time feedback of the user's click-through rate, dwell time, and sliding behavior, and optimize the generation effect of the display scheme through a feedback loop;

[0143] The hierarchical context control module, including a macro context control module, a meso context control module, and a micro context control module, is used to adjust the display scheme layer by layer according to user needs, device information, and real-time interaction behaviors;

[0144] The feedback data processing module is used to collect and process the user's interaction feedback data, input the feedback data into the dynamic context preference portrait and the adaptive neural regulation module, and iteratively update the user preference portrait based on historical interaction feedback;

[0145] The context similarity detection module is used to analyze the similarity between the current context and historical contexts based on a context similarity detection model, determine the applicability of historical display schemes in combination with a meta-learning algorithm, and transfer the historical display schemes with high applicability to the current context through a transfer learning module;

[0146] The storage module is used to record and save the generation effect and user feedback data in the current context after the transfer and fine-tuning of the display scheme, and update the context feature database.

[0147] Example 1:

[0148] To verify the feasibility of the present invention in implementation, the present invention is applied to the personalized shopping display of a large e-commerce platform. The platform has a large number of users and significant personalized needs. Different users expect to obtain different shopping contents and display methods in multiple contexts such as time, location, and device. The traditional display system of the e-commerce platform is difficult to adapt to the diverse needs of users. Usually, fixed content is displayed without considering the user's real-time context, resulting in a low matching degree between the displayed content and user needs, affecting the user's browsing experience, dwell time, and purchase conversion rate. To improve the user experience and interaction rate, the platform introduces a context-adaptive display system trained based on a display scheme data model to meet the high demands of users for personalized display.

[0149] In this embodiment, the system first collects the user's context data in real time through Internet of Things devices, including the temperature, light, device type at the user's location, and the environmental information where the user is located. The multi-head attention mechanism is used to perform weighted fusion on the context features of each dimension, and integrate this information into a single context vector. Suppose a user opens the application of an e-commerce platform through a mobile device on a weekday afternoon. The system records the user's device type, location (office area), light condition (bright), and recent browsing history. The system generates and updates the user's context preference profile, which shows that the user prefers to browse smart home and office supplies during working hours and tends to a bright and simple display style.

[0150] When generating a display plan, the system uses a conditional generative adversarial network to take this dynamic context preference profile as the input condition and generate a preliminary display plan that conforms to the user's current context. The generator generates display content including product recommendations, display layout, and color matching style, and the discriminator evaluates the fitness of the generated content to ensure that the display content meets the user's real-time needs. Suppose the generated preliminary display plan includes a group of product recommendations for promoting office supplies and smart home, with a simple and bright layout, mainly showing a white background and light-colored text. During the generation process, the system monitors the user's interaction behavior data in real time, including click-through rate, dwell time, and sliding frequency. Through the adaptive neural regulation module, the weights of the generator are continuously adjusted according to the real-time feedback to optimize the generation effect of the display content and ensure that the system can quickly respond to the changes in the user's needs.

[0151] When the user swipes and has a long dwell time on the display page, the system will adjust the display layout to highlight specific products and moderately optimize the display order of the content. In hierarchical context control, the system first macroscopically identifies the user's current demand theme as "daily necessities in the office scenario" and further refines it to the display style adaptation at the meso level (such as the bright color style during the day); at the micro level, details such as button position and product image size are adjusted in real time to ensure compliance with the user's visual preferences and interaction needs. For example, when the click-through rate of the user in selecting a desk significantly increases, the system will automatically move the recommended content of this category to the front of the page to enhance the recommendation effect.

[0152] To verify the beneficial effects of the system of the present invention, the e-commerce platform conducted a system comparison experiment in different scenarios. The comparison experiment was carried out in different locations and contexts to evaluate the advantages of the context adaptive display system in personalized recommendation. In this experiment, the system collected context features at different locations and time periods of the user, and the collected data included temperature, light, device type, and user dwell time, etc., to analyze the performance of the system in terms of real-time adaptability, matching degree, and click-through conversion rate.

[0153] The system conducted experiments in office, café, and home scenarios at different times in the morning and afternoon, and recorded the interaction data of users when using the traditional static display and the context - adaptive display system. The following are some of the collected data and analysis results.

[0154] Table 1 Comparison table of user behavior data in different usage scenarios

[0155]

[0156] The experimental results show that with the support of the context - adaptive display system, the user's stay time, click - through rate, and conversion rate in various scenarios have been significantly improved. In the office scenarios in the morning and afternoon, the average stay time of users after using the context - adaptive system has increased by about 30% compared with the traditional system, the click - through rate has increased by more than 40%, and the conversion rate has increased significantly; in café and home scenarios, whether in bright or dark environments, the context - adaptive system can adjust the display content according to the current context characteristics, thereby enhancing the user experience and interaction performance.

[0157] For example, in the home scenario with dim light, the system will preferentially display the display interface with a dark theme to reduce the interference of bright colors. In the café scenario in the evening, the system automatically optimizes the tone and layout of the display content to conform to the user's current viewing comfort. These adjustments make the browsing experience of users in the context - adaptive system more fluent, and further improve the matching degree between the display content and user needs.

[0158] Through comparative experiments, it is verified that the context - adaptive display system provided by the present invention has significant advantages in personalized display. Its context - feature acquisition and feedback mechanism quickly responds to user needs in diverse scenarios, achieving higher user stay time and click - through conversion rate, proving that the present invention can effectively solve the problem of insufficient personalized recommendation in traditional systems.

[0159] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for training a display scheme data model, characterized in that: The steps include: S1. Collect users’ micro-context data through IoT devices, extract macro-context data from user device information, and construct a multi-source context data set; S2. Preprocess the multi-source context data set, weight the context features from different sources through a multi-head attention mechanism, and integrate the multi-source context features into a single context vector; S3, extract preference features from the user's historical interaction data, generate a user preference portrait, and fuse the context vector with the user preference portrait to generate a dynamic context preference portrait; S4, inputting the generated dynamic situation preference portrait into the conditional generative adversarial network as a condition to generate a display plan that meets the current situation; S5. Embed an adaptive neural control module, use the context feature feedback loop to dynamically adjust the generation strategy of the conditional generative adversarial network, adjust the weight of the generation model in real time according to the user's click rate, dwell time and sliding behavior, and optimize the display solution generation effect; S6. Introducing a hierarchical situational control mechanism, wherein the hierarchical situational control mechanism includes macro situational control, meso situational control and micro situational control; S7, collecting feedback data of user interactions, inputting the feedback data into the dynamic situational preference profile and adaptive neural control module, adjusting the weight parameters of the conditional generative adversarial network, and iteratively updating the user preference profile based on historical interaction feedback; S8. Analyze the similarity between the current situation and the historical situation through situation similarity detection, apply the meta-learning algorithm to determine the applicability of the display scheme, and use transfer learning to transfer the effective display scheme in the historical situation to the current situation; The S6 specifically includes: S61, identify the user's current main demand context, determine the theme and overall direction of the displayed content based on the user's behavior data and preference portrait, and analyze the user's contextual needs through the context weights. Adjust the theme of the displayed content to match the user's needs and situations; S62: Adapt the content style based on the user device information and environmental factors. The meso-scenario control module applies the corresponding style template based on different device and environmental characteristics, and uses the meso-scenario style weight coefficient Controlling style transfer; S63. When the user interacts with the displayed content, the user's click, slide and stay behaviors are detected in real time, and the details of the displayed content are adjusted according to the micro-interaction behaviors. The micro-context control module dynamically updates the displayed content according to the user's real-time feedback, including adjusting the button color, content typesetting and layout, and controlling the weight coefficient through the micro-context Control the adjustment range of display details; S64. Based on the hierarchical context control mechanism, a multi-level fitness score of the displayed content is defined to measure the adaptability of the displayed content at different context levels: ； in, represents the multi-level fitness score of the displayed content, represents the macro-scenario control weight coefficient, represents the macro-scenario fitness, represents the micro-situational fitness, Indicates the macro situation control The weight parameter of each feature, Indicates the macro situation control Features, represents the number of features in the macro-context control, Indicates the first The weight parameter of each feature, Indicates the first Features, represents the number of features in the meso-situational control, Indicates the micro-situational control The weight parameter of each feature, Indicates the micro-situational control Features, represents the number of features in micro-situational control; S65, when fitness score When the preset threshold is reached, it is considered that the displayed content has achieved the expected context matching effect at the macro, meso and micro context levels, and it is determined that the displayed content meets the multi-level context needs of users; S66, if the fitness score If the preset threshold is not reached, the weight coefficients of the macro-context control module, the meso-context control module and the micro-context control module are updated through the back-propagation mechanism to gradually optimize the adaptability of the displayed content.

2. A method for training a display scheme data model according to claim 1, characterized in that: The S4 specifically includes: S41. Obtaining dynamic situational preference portraits ,in Table 1 Dynamic Situational Preference Profile feature dimension values; S42. Profiling dynamic situational preferences The input conditional generative adversarial network generator is generated by Mapped to hidden vector , and combined to generate the noise vector , generate a preliminary display content vector : ； in, represents the hyperbolic tangent activation function, represents the adaptive adjustment parameter, The weight matrix representing the dynamic situational preference profile, represents the weight matrix of the hidden layer vector, represents the weight matrix of the noise vector, represents the bias term, The number of feature dimensions representing the dynamic situational preference profile, represents the number of dimensions of the hidden layer vector, represents the number of dimensions of the noise vector, represents the smoothing term, , and Represent the weight coefficients in the weight matrix of dynamic situation preference portrait, hidden layer vector and noise vector respectively, represents the hidden layer vector feature dimension values, represents the first feature dimension values; S43, based on the preliminary display content vector outputting a preliminary display scheme, wherein the preliminary display scheme includes layout elements, color matching elements and content elements, wherein the layout elements include button positions and image display area size, the color matching elements include background color and text color, and the content elements include the type, style and text content of the display content; S44. Use the discriminator to evaluate the output of the generator, with the input condition being the dynamic situation preference portrait With the initial presentation content vector , the discriminator judges the generated solution according to the matching degree of the displayed content and outputs the matching score: ； in, represents the matching score, represents the exponential function, represents the weight matrix in the discriminator, represents the context vector weight in the discriminator, represents the bias term, Representing dynamic situational preference profiles With the initial presentation content vector The combined feature vector of Representing dynamic situational preference profiles The characteristic vector of Represents each weight item in the context vector weight, Represents the initial display content vector Each weight item in Represents the initial display content vector Each eigenvalue in represents the number of feature dimensions of the context vector, Represents the initial display content vector The number of feature dimensions; S45, if the matching score If the set matching threshold is not reached, the weight matrix and bias value of the generator are optimized for directional propagation to generate display content that is more in line with the current situation; S46. Through multiple iterations, during the alternating training of the generator and the discriminator, the display scheme generated by the generator gradually conforms to the dynamic situation preference portrait. , and finally output a display plan that suits the current situation.

3. The method for training a display scheme data model according to claim 1, characterized in that: The S5 specifically includes: S51. Collect real-time user interaction data, including click-through rate , Dwell time and sliding behavior , the click rate Indicates the user's interest in the displayed content, the dwell time Reflects the user's attention, the sliding behavior Indicates the user's tendency to explore displayed content and interaction area preference; S52. Click-through rate , Dwell time and sliding behavior Standardize and convert to a unified feedback score: ； in, represents a unified feedback score, Represents the weight coefficient of click rate, represents the weight coefficient of the residence time, Represents the weight coefficient of sliding behavior, represents the smoothing coefficient, represents the conditional generative adversarial network weight parameters for layer weights, represents the conditional generative adversarial network The eigenvalues ​​of the layer, Represents the number of layers of the generator in the conditional generative adversarial network; S53. Unify feedback scores Input to the adaptive neural control module and use a unified feedback score Adjust the weight matrix of the conditional generative adversarial network generator; S54, using a gradient descent method to update in real time the generator weight matrix, the hidden layer vector weight matrix, the noise vector weight matrix and the bias value in the conditional generative adversarial network; S55, introduce the control parameters of the adaptive feedback loop , and ,in Indicates the response sensitivity of the click rate. represents the response sensitivity of the residence time, Indicates the response sensitivity of the sliding behavior; S56. After updating and iterating the weights of the feedback loop, the display solution generation effect is optimized. When the click rate, dwell time and sliding behavior reach a predetermined threshold, it is determined that the generated display solution has met the current situation requirements.

4. The method for training a display scheme data model according to claim 1, characterized in that: The S7 specifically includes: S71. Collect comprehensive user interaction data in displayed content, including click-through rate, browsing time, sliding frequency and hot spots, to form multi-dimensional interaction data; S72, converting the collected multi-dimensional interaction data into a context feedback vector, and integrating it with the user's historical portrait data after data cleaning and preprocessing, and generating a personalized vector with time and context specificity by analyzing the user's preference fluctuations in different contexts; S73, the dynamic situation preference portrait module generates a multi-stage user portrait update strategy based on the input of the situation feedback vector, wherein the update strategy separates short-term preferences from long-term preferences based on the user's interaction frequency, time interval, and changes in situational needs, and constructs a preference feature vector that meets the user's recent interests; S74, the adaptive neural control module receives the updated situation preference portrait information and automatically detects the deviation area of ​​the user's interest characteristics; S75. Through the multi-level feedback control mechanism of the adaptive neural control module, preference adjustment weights are further introduced to enable the generator to automatically adjust the display scheme of the displayed content. The preference adjustment weights are based on the user context feedback frequency and context change trend to dynamically adjust the style, color matching and structural display characteristics of the generated content. S76. In multiple rounds of iterative optimization, by recording the generated effect after each user interaction, gradually build a historical preference map in the user portrait, accumulate the user's preference change trajectory, identify the long-term pattern of user preference based on the historical preference map generator, and give priority to the long-term pattern of user preference when generating content.

5. The method for training a display scheme data model according to claim 1, characterized in that: The S8 specifically includes: S81. Collect current user context data, including the user's real-time behavior characteristics, device information, and environmental parameters, to form a current context feature vector ; S82. Calculate the current situation feature vector based on the situation similarity detection model and the historical context feature vector set The similarity between them is calculated based on the cosine similarity or Euclidean distance measurement method to select the feature vector that is most similar to the current situation; S83. Apply the meta-learning algorithm to determine the adaptability of the display scheme in the historical context to the current context: ； in, represents the suitability score, represents the number of historical context features, Indicates The weight of the feature, Indicates The weight of the feature, Indicates the current situation eigenvalues, The historical context eigenvalues, represents the exponential function, represents the exponential decay coefficient, represents the smoothing coefficient, represents the nonlinear adjustment coefficient, Indicates the current situation eigenvalues, The historical context eigenvalues; S84. For the display scheme determined to be applicable, the transfer learning module transfers the generation model parameters in the historical context to the current context, and the transfer process includes the selection of key parameters, the conversion of feature mapping and the optimization of content features; S85. For the migrated display scheme, the generated model is fine-tuned. The fine-tuning process is based on the differences between the current context and the historical context, and the display strategy, color matching style and interaction details in the model are adjusted; S86. After completing the migration and fine-tuning, the generation effect and user feedback data in the current situation are recorded and saved as new historical situation features.

6. A display scheme data model training system, which executes the display scheme data model training method according to any one of claims 1 to 5, characterized in that: Includes the following modules: The data collection module is used to collect the user's micro-context data through the IoT device and extract the macro-context data from the user's device information to generate a context data set containing multi-source context features; The data preprocessing module is used to clean, standardize and denoise the context data set, and to perform weighted fusion of context features from different sources through a multi-head attention mechanism to generate a single context vector; The user portrait generation module is used to extract preference features from the user's historical interaction data to generate a user preference portrait, and fuse the context vector with the user preference portrait to generate a dynamic context preference portrait; The conditional generative adversarial network module includes a generator and a discriminator, which is used to receive the dynamic situation preference portrait as a conditional input and generate a display plan that conforms to the current situation; Adaptive neural control module, which is used to dynamically adjust the generation strategy of the conditional generative adversarial network based on real-time feedback from users’ click rate, dwell time, and sliding behavior, and optimize the generation effect of the display solution through feedback loop; The hierarchical context control module includes a macro context control module, a meso context control module, and a micro context control module, which is used to adjust the display scheme layer by layer according to user needs, device information, and real-time interactive behavior, including: Identify the user's current main demand context, determine the theme and overall direction of the displayed content based on user behavior data and preference portraits, and analyze the user's contextual needs through the context weights. Adjust the theme of the displayed content to match the user's needs and situations; The content style is adapted and adjusted based on the user device information and environmental factors. The meso-context control module applies the corresponding style template based on different device and environmental characteristics, and uses the meso-context style weight coefficient Controlling style transfer; When users interact with the displayed content, the user's click, slide and stay behaviors are detected in real time, and the details of the displayed content are adjusted according to the micro-interaction behaviors. The micro-context control module dynamically updates the displayed content according to the user's real-time feedback, including adjusting the button color, content typesetting and layout, and controlling the weight coefficient through the micro-context Control the adjustment range of display details; According to the hierarchical context control mechanism, a multi-level fitness score for display content is defined to measure the adaptability of display content at different context levels: ； in, represents the multi-level fitness score of the displayed content, represents the macro-scenario control weight coefficient, represents the macro-scenario fitness, represents the micro-situational fitness, Indicates the macro situation control The weight parameter of each feature, Indicates the macro situation control Features, represents the number of features in macro-context control, Indicates the first The weight parameter of each feature, Indicates the first Features, represents the number of features in the meso-situational control, Indicates the micro-situational control The weight parameter of each feature, Indicates the micro-situational control Features, represents the number of features in micro-situational control; When the fitness score When the preset threshold is reached, it is considered that the displayed content has achieved the expected context matching effect at the macro, meso and micro context levels, and it is determined that the displayed content meets the multi-level context needs of users; If the fitness score If the preset threshold is not reached, the weight coefficients of the macro context control module, the meso context control module and the micro context control module are updated through the back-propagation mechanism to gradually optimize the adaptability of the displayed content; Feedback data processing module, used to collect and process user interaction feedback data, input the feedback data into the dynamic situation preference profile and adaptive neural control module, and iteratively update the user preference profile based on historical interaction feedback; The context similarity detection module is used to analyze the similarity between the current context and the historical context based on the context similarity detection model, determine the applicability of the historical display scheme in combination with the meta-learning algorithm, and transfer the highly applicable historical display scheme to the current context through the transfer learning module; The storage module is used to record and save the generation effect and user feedback data in the current situation after completing the migration and fine-tuning of the display plan, and update the situation feature database.

Images