A machine learning-based adaptive optimization method and system for programming interface layout

Abstract

This invention discloses a machine learning-based adaptive optimization method and system for programming interface layout, comprising: acquiring multi-source data of the user in the programming interface; preprocessing the multi-source data to extract feature data and generating feature vectors for training a machine learning model; training an interface layout prediction model using a reinforcement learning algorithm on the feature vectors to obtain optimal interface layout data predicted by the current user behavior and device status; adjusting the optimal interface layout data into visual layout adjustment instructions and dynamically updating the arrangement, size ratio, and hierarchical relationship of each functional module in the programming interface; and performing online incremental learning of the interface layout prediction model by receiving user feedback information to adaptively adjust the programming interface layout. This method can intelligently identify user behavior characteristics and personalized preferences, continuously adapt to changes in user behavior and environmental evolution, and significantly improve development efficiency, interface interaction experience, and user satisfaction.

Description

Technical Field

[0001] This invention relates to the field of front-end interface development technology, specifically to a method and system for adaptive optimization of programming interface layout based on machine learning. Background Technology

[0002] In front-end development, menus are a crucial component of user interaction, and their display directly impacts user experience. With the development of mobile internet and multiple terminal devices, front-end interfaces need to adapt to various screen resolutions, such as 1024×768px, 1366×768px, and 1920×1080px. Simultaneously, users' demands for personalized interface settings are increasing. Traditional menu bar layouts often employ fixed column numbers or static spacing strategies, making it difficult to adapt to scenarios where the number and size of icons and text content dynamically change. Existing technologies suffer from the following problems:

[0003] (1) When icons and text are mixed in the layout, the width calculation is complicated and overflow or blank areas are likely to occur. Since the actual width of the combined icons and text items is inconsistent, the traditional fixed column width or evenly distributed layout method is difficult to allocate space reasonably, resulting in some menu items not being fully displayed or having too much white space.

[0004] (2) The mixed arrangement of large and small icons lacks adaptive rules and requires manual adjustment of the number of columns. The visual center of gravity and space occupied by icons of different sizes are quite different. The existing layout method cannot automatically adjust the arrangement rules according to the icon size and often relies on manual intervention to ensure visual consistency.

[0005] (3) The impact of text length on column width was not fully considered, resulting in layout misalignment and frequent changes in text content length. Especially in systems that support multiple languages ​​or user-defined tags, ignoring text length can lead to unreasonable column width, which in turn can cause problems such as line break confusion and layout collapse. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for adaptive optimization of programming interface layout based on machine learning, in order to solve the technical problem that the existing editing interface cannot achieve automated column number adaptation and space utilization optimization, resulting in problems such as menu overflow, line break misalignment, and wasted blank areas when different devices or content change, which affect the aesthetics of the interface and the efficiency of operation.

[0007] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution:

[0008] A first aspect of the present invention provides a machine learning-based adaptive optimization method for programming interface layout, comprising the following steps:

[0009] The system acquires multi-source data from the user's programming interface, preprocesses the multi-source data to extract feature data, and generates feature vectors for training machine learning models.

[0010] The feature vectors are used to train an interface layout prediction model to obtain the optimal interface layout data for predicting the current user behavior and device status.

[0011] The optimal interface layout data is adjusted into a visual layout adjustment command, and the arrangement, size ratio and hierarchical relationship of each functional module in the programming interface are dynamically updated.

[0012] The interface layout prediction model is incrementally learned online by receiving user feedback information, and the programmed interface layout is adaptively adjusted.

[0013] As a preferred embodiment of the present invention, the method involves acquiring multi-source data from the user's programming interface, preprocessing the multi-source data to extract feature data, and generating feature vectors for training machine learning models, including:

[0014] User interaction data in the programming interface is collected as multi-source data. After data cleaning, the multi-source data is normalized. The normalized multi-source data is then encoded to obtain encoded data.

[0015] Extract the temporal features, spatial distribution features, and interaction frequency features of the encoded data to obtain feature data;

[0016] Based on the task type and contextual features on the current editing interface, the feature data is dynamically updated using a fixed time window sliding method to construct a feature vector.

[0017] As a preferred embodiment of the present invention, an information flow management set U = {T, W, E, S} is established for the feature vector in the current editing interface, where T = {t1, t2, ..., t n} represents the set of tasks in the current editing interface during the running phase. The interface information to be scheduled depends on the specific task operations in the running tasks; W = {w1, w2, ..., w n} represents the set of work locations in the current editing interface system that have established mutual trust relationships; E represents the current editing interface information flow workflow engine, which assists users in analyzing and understanding the business process in the editing interface, and generates information delegation, supervision, and interface evaluation information; S represents the editing interface information queue;

[0018] Establish a scheduling queue for the current editing interface based on the information flow management set U, and generate a scheduling execution information flow requirement table for the current editing interface;

[0019] Establish the priority of tasks in the current editing interface during the running phase, and sort the task set T, where the priority is Priority(t1)>Priority(t2)>…>Priority(t3) n ), where n represents the number of tasks in the current editing stage;

[0020] A decision model is established based on the task requirements in the current editing interface, and the scheduling data of the current editing interface is obtained.

[0021] In a preferred embodiment of the present invention, the decision model uses task requirements as the basis for adjusting the interface layout, forming scheduling data with information flow node information, including:

[0022] A decision knowledge base is constructed by analyzing the task requirements of the current editing interface, and a task queue is constructed based on the task requirements.

[0023] The task queue is driven by the current editing interface information flow workflow engine to analyze task requirements, adjust the information layout of the interface, and generate information flow data for the editing interface.

[0024] The task execution status in the information flow data is analyzed, the tasks are decomposed and fed back to the information flow workflow engine of the current editing interface, and the editing interface is scheduled and adjusted to form scheduling data with information flow node information.

[0025] As a preferred embodiment of the present invention, a reinforcement learning algorithm is used to train an interface layout prediction model on the scheduling data to obtain the optimal interface layout data for predicting the current user behavior and device status, including:

[0026] Obtain user behavior features and device status parameters contained in the scheduling data, and encode the user behavior features and device status parameters into a dynamic state vector;

[0027] The arrangement mode, spatial priority weight, and resource distribution strategy of the interface components in the current editing interface are used as a set of adjustable actions to make multi-dimensional layout adjustments to the current editing interface.

[0028] Based on the task completion time in the task set T and the hotspot coverage in the current editing interface, a user experience quantitative index is constructed, and combined with the device status parameters, a system load balancing coefficient is calculated to construct an interface layout prediction model.

[0029] The interface layout prediction model is trained using a near-end strategy optimization algorithm combined with historical interaction data. The convergence optimal strategy of the interface layout prediction model is calculated through a composite reward function to obtain the optimal interface layout data that constrains the current editing interface.

[0030] As a preferred embodiment of the present invention, the near-end strategy optimization algorithm obtains the importance of the task information of the current editing interface by evaluating the coverage of click hotspots in the current editing interface, including:

[0031] Based on indicators such as the hotspot areas, click hotspot coverage, and user behavior characteristics of the current editing interface, the importance of the task information of the current editing interface is evaluated, and its expression is:

[0032]

[0033] Among them, fn j This represents the number of clicks within the j-th hotspot area. ω(fn) represents the total number of clicks made by the user during the current editing process in the editing interface. j This indicates the importance of the j-th hotspot area based on user behavior characteristics.

[0034] As a preferred embodiment of the present invention, a particle swarm optimization algorithm based on global search is used to construct an interface layout prediction model to seek the optimal solution, including:

[0035] The optimal interface layout data is initialized, an initial population is randomly generated, and the user experience quantification index is taken as a collective of m particles, with the particle number set.

[0036] The interface layout prediction model is trained by combining historical interaction data. All particles are diffused at a rate v in the D-dimensional search space. The particle swarm optimization process is performed to update the velocity and position of each particle.

[0037] A linearly decreasing weighting method is used to introduce inertia weights for global search of the particle swarm. The number of iterations of the particle swarm changes as follows:

[0038]

[0039] Where φ1 represents the initial inertia weight value, φ0 represents the inertia weight value after the particle swarm algorithm has iterated to the maximum number of times, and I max I represents the maximum number of iterations, and I represents the current number of iterations.

[0040] By examining the historical optimal points of all particles, the optimal path is obtained, and the coverage of the click hotspot and the task completion time in the current editing interface are adjusted.

[0041] As a preferred embodiment of the present invention, the optimal interface layout data is adjusted into visual layout adjustment instructions, and the arrangement, size ratio, and hierarchical relationship of each functional module in the programming interface are dynamically updated, including:

[0042] The optimal particle position vector output by the interface layout prediction model is parsed into specific interface parameter configurations, and formatted and encapsulated through a preset layout instruction protocol to generate visual layout script code that can be executed on the front end.

[0043] A visual rendering scheduling module is integrated into the programming interface system. It receives layout adjustment instructions from the layout prediction model and maps them to the corresponding functional module rendering actions.

[0044] Based on the position information in the layout adjustment instructions, the functional modules are rearranged in an instant switching manner to maintain the visual coherence of the overall interface and the continuity of user operation.

[0045] Based on the user's current interaction context, the display level and focus of each module are set, and a user confirmation feedback channel is set. If the user manually adjusts or triggers a rollback operation, the current adjustment behavior is recorded as a training sample and synchronously fed back to the interface layout prediction model.

[0046] As a preferred embodiment of the present invention, the interface layout prediction model is incrementally learned online by receiving user feedback information, and the programmed interface layout is adaptively adjusted, including:

[0047] Real-time collection of user feedback behavior data on the current layout; cleaning and normalization of the collected feedback data; and combining it with the original input feature vector to form a complete feedback sample.

[0048] The incremental random forest algorithm is used to add the feedback samples one by one into the training process, fine-tuning the key parameters of the interface layout prediction model and outputting a new optimal layout scheme.

[0049] Based on the optimal layout scheme, a visual layout adjustment command is triggered to automatically optimize the position, size, and hierarchy of each functional module in the programming interface, thereby achieving continuous adaptive evolution of the interface layout.

[0050] A machine learning-based adaptive optimization system for programming interface layout, comprising:

[0051] The multi-source data acquisition module is used to collect multi-source interactive behavior data of users in the programming interface in real time;

[0052] The feature extraction and vectorization module cleans, normalizes, and encodes the collected multi-source data, extracts user behavior features, interface status features, and personalized preference features, and constructs standardized feature vectors as input samples for machine learning models.

[0053] The layout prediction modeling module constructs an interface layout prediction model based on the particle swarm optimization algorithm. It takes historical user behavior data and current feature vectors as input and searches for the optimal interface configuration scheme in a high-dimensional layout parameter space.

[0054] The visualization instruction generation module parses the optimal layout data into visual layout adjustment instructions that can be executed on the front end, describes the interface components in JSON format, and passes them to the interface control layer through the rendering engine interface.

[0055] The interface dynamic update module dynamically updates the arrangement, size ratio, and display hierarchy of each functional module in the programming interface based on the received layout adjustment instructions.

[0056] Compared with the prior art, the present invention has the following advantages:

[0057] This invention employs a parallel algorithm to divide video frame data into regions using a Tile divider. Frame numbers are added to video frames within each Tile region to facilitate matching the source video position during video decoding, preventing duplicate or missing frames and enhancing the user's immersion and realism.

[0058] By utilizing a rendering server to schedule video frame rendering tasks in a tile area and then scheduling these tasks through the GPU computing architecture of the server cluster, the system achieves optimal cluster computing power to enable larger-scale rendering and acceleration tasks. This approach is suitable for scenarios where the computational resource requirements of rendering tasks exceed the computing power of a single GPU. Furthermore, the system employs an internal distribution mechanism within the rendering server, enabling fast and efficient game management. The rendering server does not require external storage, reducing deployment costs, and supports rapid reset, achieving efficient resource utilization and enhancing user immersion and realism. Attached Figure Description

[0059] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0060] Figure 1 A flowchart of a machine learning-based adaptive optimization method for programming interface layout provided in an embodiment of the present invention;

[0061] Figure 2 A block diagram of a machine learning-based programming interface layout adaptive optimization system provided in an embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] like Figure 1 and Figure 2 As shown, this invention provides a machine learning-based adaptive optimization method for programming interface layout, comprising the following steps:

[0064] The system acquires multi-source data from the user's programming interface, preprocesses the multi-source data to extract feature data, and generates feature vectors for training machine learning models.

[0065] In this embodiment, by collecting multi-source behavioral data of users in the programming interface, such as mouse trajectory, window dwell time, code editing mode, etc., and extracting feature vectors, and combining them with reinforcement learning algorithms to train the interface layout prediction model, it is possible to accurately identify the user's operating habits and current task status, thereby predicting and generating the interface layout scheme that best meets the user's needs, improving user experience and development efficiency.

[0066] The feature vectors are used to train an interface layout prediction model to obtain the optimal interface layout data for predicting the current user behavior and device status.

[0067] The optimal interface layout data is adjusted into a visual layout adjustment command, and the arrangement, size ratio and hierarchical relationship of each functional module in the programming interface are dynamically updated.

[0068] In this embodiment, the optimal interface layout data is converted into visual layout adjustment instructions, which can dynamically adjust the arrangement, size ratio and hierarchical relationship of each functional module in the programming interface in real time, ensuring that the interface changes smoothly and naturally and responds quickly, avoiding operation interruption and inefficiency caused by manual configuration.

[0069] The interface layout prediction model is incrementally learned online by receiving user feedback information, and the programmed interface layout is adaptively adjusted.

[0070] In this embodiment, the interface layout prediction model is incrementally learned online by receiving user feedback information, enabling the system to continuously adapt to changes in user behavior and preferences, forming a personalized long-term optimization path, enhancing the intelligence and sustainability of the system, and thus improving the human-computer interaction experience.

[0071] Acquire multi-source data from the user's programming interface, preprocess the multi-source data to extract feature data, and generate feature vectors for training machine learning models, including:

[0072] User interaction data in the programming interface is collected as multi-source data. After data cleaning, the multi-source data is normalized. The normalized multi-source data is then encoded to obtain encoded data.

[0073] In this embodiment, multi-source data is processed to achieve systematic collection and structured processing of multi-source data, thereby improving the accuracy of feature extraction.

[0074] Extract the temporal features, spatial distribution features, and interaction frequency features of the encoded data to obtain feature data;

[0075] Based on the task type and contextual features on the current editing interface, the feature data is dynamically updated using a fixed time window sliding method to construct a feature vector.

[0076] In this embodiment, features including temporal features, spatial distribution features, and interaction frequency features are extracted to construct a feature set that comprehensively reflects user behavior patterns, enabling machine learning models to more accurately identify user habits and intentions. The temporal features include the user's operation sequence and rhythm, the spatial distribution features include click hotspots and window layout preferences, and the interaction frequency features refer to the user's high-frequency operation areas.

[0077] In this embodiment, the constructed feature vector has a unified structure and clear semantics. It is not only applicable to the current interface layout prediction task, but can also be used as input for other downstream tasks such as intelligent recommendation and user behavior analysis, and has good scalability and reusability.

[0078] In this embodiment, a fixed time window sliding method is used to dynamically update the feature data, ensuring that the model always makes predictions based on the latest user behavior, which has good timeliness and adaptability and helps to capture the changing trends of user behavior preferences.

[0079] In the current editing interface, an information flow management set U = {T, W, E, S} is established for the feature vector, where T = {t1, t2, ..., t...} n} represents the set of tasks in the current editing interface during the running phase. The interface information to be scheduled depends on the specific task operations in the running tasks; W = {w1, w2, ..., w n} represents the set of work locations in the current editing interface system that have established mutual trust relationships; E represents the current editing interface information flow workflow engine, which assists users in analyzing and understanding the business process in the editing interface, and generates information delegation, supervision, and interface evaluation information; S represents the editing interface information queue;

[0080] In this embodiment, by establishing an information flow management set for feature vectors in the current editing interface, including task sets, work location trust relationships, information flow engine, and information queue, a comprehensive model of the task status during the interface operation phase is achieved. This mechanism helps to unify scheduling logic, optimize resource allocation, and provide a structured data foundation for intelligent decision-making.

[0081] Establish a scheduling queue for the current editing interface based on the information flow management set U, and generate a scheduling execution information flow requirement table for the current editing interface;

[0082] In this embodiment, a scheduling execution information flow requirement table is automatically generated based on the information flow management set. This allows for dynamic adjustment of the scheduling strategy according to the current task status, ensuring that the scheduling of interface resources is highly consistent with the user's operational intent, thereby improving the system's response efficiency and user experience.

[0083] Establish the priority of tasks in the current editing interface during the running phase, and sort the task set T, where the priority is Priority(t1)>Priority(t2)>…>Priority(t3) n ), where n represents the number of tasks in the current editing stage;

[0084] In this embodiment, the task set T is sorted according to a set priority, so that the system can prioritize responding to the core tasks that users care about, avoid low-priority tasks from interfering with the user's operation process, and effectively improve development efficiency and the rationality of task execution.

[0085] A decision model is established based on the task requirements in the current editing interface, and the scheduling data of the current editing interface is obtained.

[0086] In this embodiment, by establishing a task-driven decision-making model, the system can combine historical behavior with current context information to predict the optimal scheduling path and dynamically adjust resource allocation, further improving the intelligence and automation level of interface layout and function calls.

[0087] In this embodiment, the information flow engine not only helps users understand the business process, but also supports the generation of information delegation, supervision and interface evaluation information, enabling users to better grasp the interface operation status and enhance the controllability and transparency of human-computer interaction.

[0088] The decision-making model uses task requirements as the basis for adjusting the interface layout, forming scheduling data with information flow node information, including:

[0089] A decision knowledge base is constructed by analyzing the task requirements of the current editing interface, and a task queue is constructed based on the task requirements.

[0090] In this embodiment, the decision model, by constructing a decision knowledge base and combining it with a task queue mechanism, can dynamically adjust the interface layout based on the semantic information of the user's current editing task, making the content displayed on the interface more in line with the actual operation intention, and significantly improving interaction efficiency and user experience.

[0091] The task queue is driven by the current editing interface information flow workflow engine to analyze task requirements, adjust the information layout of the interface, and generate information flow data for the editing interface.

[0092] In this embodiment, the information flow workflow engine is used to drive the task queue to perform in-depth analysis of task requirements and provide real-time feedback to the interface layout adjustment process, forming a closed-loop information processing mechanism, which improves the coordination and consistency between task execution and interface response.

[0093] The task execution status in the information flow data is analyzed, the tasks are decomposed and fed back to the information flow workflow engine of the current editing interface, and the editing interface is scheduled and adjusted to form scheduling data with information flow node information.

[0094] In this embodiment, by analyzing the task execution status in the information flow data in real time, the system can automatically break down complex tasks into sub-tasks and feed them back to the information flow engine for rescheduling. This not only enhances the manageability of tasks but also helps to improve the orderliness and transparency of the development process.

[0095] The scheduling data is used to train an interface layout prediction model using a reinforcement learning algorithm to obtain the optimal interface layout data for predicting the current user behavior and device state, including:

[0096] Obtain user behavior features and device status parameters contained in the scheduling data, and encode the user behavior features and device status parameters into a dynamic state vector;

[0097] In this embodiment, by extracting user behavior features and device status parameters from scheduling data and encoding them into a unified dynamic state vector, it is possible to comprehensively reflect the user's operating intentions, usage habits, and the current device operating environment, providing a high-dimensional and semantically rich input foundation for the interface layout prediction model.

[0098] The arrangement mode, spatial priority weight, and resource distribution strategy of the interface components in the current editing interface are used as a set of adjustable actions to make multi-dimensional layout adjustments to the current editing interface.

[0099] In this embodiment, the arrangement mode of interface components, spatial priority weights, resource distribution strategies, etc. are used as the action space of reinforcement learning, enabling the system to finely control the interface layout in multiple dimensions, taking into account both visual aesthetics and functional usability, and meeting the personalized needs in complex interaction scenarios.

[0100] Based on the task completion time in the task set T and the hotspot coverage in the current editing interface, a user experience quantitative index is constructed, and combined with the device status parameters, a system load balancing coefficient is calculated to construct an interface layout prediction model.

[0101] In this embodiment, a quantitative indicator of user experience is constructed based on task completion time and click hotspot coverage, enabling the system to optimize the interface in a "user-centric" manner, avoiding the subjectivity and one-sidedness of traditional rule-based layout, and enhancing the goal orientation and evaluation objectivity of model training.

[0102] The interface layout prediction model is trained using a near-end strategy optimization algorithm combined with historical interaction data. The convergence optimal strategy of the interface layout prediction model is calculated through a composite reward function to obtain the optimal interface layout data that constrains the current editing interface.

[0103] In this embodiment, the system load balancing coefficient is calculated by integrating device status parameters into the interface layout prediction model. This ensures that interface adjustments not only consider user experience but also system performance and resource allocation efficiency, effectively preventing interface lag or response delays caused by excessive resource consumption.

[0104] In this embodiment, a near-end policy optimization algorithm is used for model training, and the policy is updated by combining the user's historical interaction data. This allows the model to maintain the stability and generalization ability of the training process while exploring new layout strategies, significantly improving the model's convergence efficiency and practical deployment feasibility.

[0105] The near-end strategy optimization algorithm assesses the importance of task information in the current editing interface by evaluating the coverage of click hotspots in the current editing interface, including:

[0106] Based on indicators such as the hotspot areas, click hotspot coverage, and user behavior characteristics of the current editing interface, the importance of the task information of the current editing interface is evaluated, and its expression is:

[0107]

[0108] Among them, fn j This represents the number of clicks within the j-th hotspot area. ω(fn) represents the total number of clicks made by the user during the current editing process in the editing interface. j This indicates the importance of the j-th hotspot area based on user behavior characteristics.

[0109] In this embodiment, the task importance evaluation expression not only considers the ratio between the number of clicks and the total number of clicks, but also introduces importance weights based on user behavior characteristics, so that the scoring results can more accurately reflect the user's personalized operation preferences and task focus, and provide more targeted training signals for reinforcement learning strategies.

[0110] A particle swarm optimization algorithm based on global search is used to construct an interface layout prediction model to seek the optimal solution, including:

[0111] The optimal interface layout data is initialized, an initial population is randomly generated, and the user experience quantification index is taken as a collective of m particles, with the particle number set.

[0112] In this embodiment, by abstracting the interface layout configuration as the position vector of particles in a D-dimensional search space and using the particle swarm optimization algorithm for global search, the potential optimal solution in the complex multi-dimensional layout space can be effectively explored, significantly improving the efficiency and accuracy of automatic interface layout optimization.

[0113] The interface layout prediction model is trained by combining historical interaction data. All particles are diffused at a rate v in the D-dimensional search space. The particle swarm optimization process is performed to update the velocity and position of each particle.

[0114] In this embodiment, user behavior data such as task completion time and click hotspot coverage are transformed into the fitness evaluation criteria of particle swarm optimization, ensuring that each iteration converges in the direction of improving user experience and avoiding the problem of "visually beautiful but impractical" in traditional layout optimization.

[0115] A linearly decreasing weighting method is used to introduce inertia weights for global search of the particle swarm. The number of iterations of the particle swarm changes as follows:

[0116]

[0117] Where φ1 represents the initial inertia weight value, φ0 represents the inertia weight value after the particle swarm algorithm has iterated to the maximum number of times, and I max I represents the maximum number of iterations, and I represents the current number of iterations.

[0118] In this embodiment, a linearly decreasing inertial weight strategy is adopted to balance global search and local convergence capabilities. This ensures that the algorithm maintains strong global search capabilities in the early stages and gradually focuses on local optimal solutions in the later stages, thereby improving the stability and convergence speed of the algorithm.

[0119] By examining the historical optimal points of all particles, the optimal path is obtained, and the coverage of the click hotspot and the task completion time in the current editing interface are adjusted.

[0120] In this embodiment, historical user interaction data is integrated during particle swarm optimization, enabling the model to learn the operational preferences of different users and generate an interface layout scheme that better suits their usage habits, thereby improving the intelligence and personalization of the system.

[0121] The optimal interface layout data is adjusted into visual layout adjustment instructions, and the arrangement, size ratio, and hierarchical relationship of each functional module in the programming interface are dynamically updated, including:

[0122] The optimal particle position vector output by the interface layout prediction model is parsed into specific interface parameter configurations, and formatted and encapsulated through a preset layout instruction protocol to generate visual layout script code that can be executed on the front end.

[0123] In this embodiment, the optimal particle position vector output by the interface layout prediction model is parsed into specific interface parameter configurations, such as window coordinates, size, and hierarchy, and encapsulated into visual layout script code that can be executed on the front end, such as JSON instructions and CSSGrid layout statements. This bridges the last mile from intelligent prediction to practical application, enabling machine learning results to truly be implemented in the user interface.

[0124] A visual rendering scheduling module is integrated into the programming interface system. It receives layout adjustment instructions from the layout prediction model and maps them to the corresponding functional module rendering actions.

[0125] In this embodiment, a visual rendering scheduling module is integrated into the programming interface system. This module can receive and parse layout adjustment instructions from the model and map them to corresponding functional module rendering actions. This module supports mainstream development frameworks such as Electron, WebComponents, and Qt, improving the system's cross-platform adaptability and deployment flexibility.

[0126] Based on the position information in the layout adjustment instructions, the functional modules are rearranged in an instant switching manner to maintain the visual coherence of the overall interface and the continuity of user operation.

[0127] In this embodiment, the functional modules are rearranged in real-time switching mode to ensure that users do not experience visual discontinuity or operational interruption during the interface update process, thereby improving the overall smoothness and usability of the interaction.

[0128] Based on the user's current interaction context, the display level and focus of each module are set, and a user confirmation feedback channel is set. If the user manually adjusts or triggers a rollback operation, the current adjustment behavior is recorded as a training sample and synchronously fed back to the interface layout prediction model.

[0129] In this embodiment, a user confirmation feedback channel is set up. If the user manually adjusts or triggers a rollback operation, the system will automatically record the behavior as a new training sample and synchronously feed it back to the interface layout prediction model for subsequent online incremental learning. This enables the system to have continuous evolution capabilities. By combining user historical preferences, current task type and real-time feedback, the system can generate highly personalized interface layout schemes, avoiding a "one-size-fits-all" layout strategy and significantly improving the consistency and satisfaction of experience for different user groups and different usage scenarios.

[0130] By receiving user feedback, the interface layout prediction model is incrementally learned online to adaptively adjust the programmed interface layout, including:

[0131] Real-time collection of user feedback behavior data on the current layout; cleaning and normalization of the collected feedback data; and combining it with the original input feature vector to form a complete feedback sample.

[0132] In this embodiment, by collecting real-time user feedback behavior data on the current interface layout, such as manually adjusting the window size, back operation, and changes in click hotspots, and structuring this data into training samples, the system can continuously correct and optimize the model prediction results based on the actual usage of users, significantly improving user experience and satisfaction.

[0133] The incremental random forest algorithm is used to add the feedback samples one by one into the training process, fine-tuning the key parameters of the interface layout prediction model and outputting a new optimal layout scheme.

[0134] In this embodiment, the collected feedback data is cleaned and normalized to eliminate outliers and dimensional differences, avoid noise interference, and ensure the accuracy and consistency of the feedback samples. After each new feedback sample is received, the system automatically fine-tunes the key parameters of the interface layout prediction model, such as the weight matrix and decision node priority, so that the model gradually adapts to the user's personalized usage habits and improves the accuracy and fit of the interface recommendation.

[0135] Based on the optimal layout scheme, a visual layout adjustment command is triggered to automatically optimize the position, size, and hierarchy of each functional module in the programming interface, thereby achieving continuous adaptive evolution of the interface layout.

[0136] In this embodiment, based on the optimal layout scheme, the system automatically triggers a visual layout adjustment command to dynamically update the position, size, and hierarchical relationship of each functional module in the programming interface, ensuring that the interface is always in the optimal state and has the ability to continuously evolve by becoming "smarter the more it is used".

[0137] The second embodiment provides a machine learning-based adaptive optimization system for programming interface layout, comprising:

[0138] The multi-source data acquisition module is used to collect multi-source interactive behavior data of users in the programming interface in real time;

[0139] The feature extraction and vectorization module cleans, normalizes, and encodes the collected multi-source data, extracts user behavior features, interface status features, and personalized preference features, and constructs standardized feature vectors as input samples for machine learning models.

[0140] The layout prediction modeling module constructs an interface layout prediction model based on the particle swarm optimization algorithm. It takes historical user behavior data and current feature vectors as input and searches for the optimal interface configuration scheme in a high-dimensional layout parameter space.

[0141] The visualization instruction generation module parses the optimal layout data into visual layout adjustment instructions that can be executed on the front end, describes the interface components in JSON format, and passes them to the interface control layer through the rendering engine interface.

[0142] The interface dynamic update module dynamically updates the arrangement, size ratio, and display hierarchy of each functional module in the programming interface based on the received layout adjustment instructions.

[0143] In this embodiment, user interaction behavior data is acquired in real time by collecting multi-source data, providing rich and authentic behavioral evidence, improving the understanding of user operation intentions and the data foundation of the prediction model. The original data is cleaned, normalized and encoded by the feature extraction and vectorization module to construct feature vectors in a unified format, which not only improves data availability, but also provides good support for transfer learning and model reuse in different task scenarios. The layout prediction modeling module is constructed using the particle swarm optimization algorithm, which can quickly locate the optimal interface configuration scheme in the high-dimensional layout parameter space. It not only takes into account the global search capability and local convergence efficiency, but also finds high-quality layout strategies in complex interface environments.

[0144] This invention, through the collaborative work of modules such as multi-source data acquisition, feature extraction and vectorization, layout prediction modeling, visualization instruction generation, and dynamic interface updates, can intelligently identify user behavior characteristics and personalized preferences, dynamically generate optimal interface layout schemes, and enable the interface layout prediction model to continuously adapt to changes in user behavior and environmental evolution, adjusting the arrangement, size ratio, and display level of interface components in real time. This ensures a good visual experience and operational continuity under different screen resolutions and usage scenarios, possesses continuous optimization and personalized adaptation capabilities, significantly improves development efficiency, interface interaction experience, and user satisfaction, and has good intelligence and versatility.

[0145] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for adaptive optimization of programming interface layout based on machine learning, characterized in that, Includes the following steps: The system acquires multi-source data from the user's programming interface, preprocesses the multi-source data to extract feature data, and generates feature vectors for training machine learning models. The feature vectors are used to train an interface layout prediction model to obtain the optimal interface layout data for predicting the current user behavior and device status. The optimal interface layout data is adjusted into a visual layout adjustment command, and the arrangement, size ratio and hierarchical relationship of each functional module in the programming interface are dynamically updated. The interface layout prediction model is incrementally learned online by receiving user feedback information, and the programmed interface layout is adaptively adjusted. Acquire multi-source data from the user's programming interface, preprocess the multi-source data to extract feature data, and generate feature vectors for training machine learning models, including: User interaction data in the programming interface is collected as multi-source data. After data cleaning, the multi-source data is normalized. The normalized multi-source data is then encoded to obtain encoded data. Extract the temporal features, spatial distribution features, and interaction frequency features of the encoded data to obtain feature data; Based on the task type perceived context features on the current editing interface, dynamic updates are performed using a fixed time window sliding method to construct feature vectors from the feature data; An information flow management set is established for the feature vector in the current editing interface. ,in, This represents the set of tasks in the current editing interface at the current stage. The interface information to be scheduled depends on the specific task operations within the running tasks. This represents the set of work locations in the current editing interface system that have established mutual trust relationships; This represents the current editing interface information flow workflow engine, assisting users in analyzing and understanding the business process within the editing interface, and generating information delegation, supervision, and interface evaluation information; This represents the information queue in the editing interface; According to the information flow management set Establish a scheduling queue for the current editing interface and generate a scheduling execution information flow requirement table for the current editing interface; Establish the priority of tasks in the current editing interface during the running phase, and sort the task set T, where the priority is... , where n represents the number of tasks in the current editing stage; Establish a decision model based on the task requirements in the current editing interface, and obtain the scheduling data of the current editing interface; The decision-making model uses task requirements as the basis for adjusting the interface layout, forming scheduling data with information flow node information, including: A decision knowledge base is constructed by analyzing the task requirements of the current editing interface, and a task queue is constructed based on the task requirements. The task queue is driven by the current editing interface information flow workflow engine to analyze task requirements, adjust the information layout of the interface, and generate information flow data for the editing interface. The task execution status in the information flow data is analyzed, the tasks are decomposed and fed back to the information flow workflow engine of the current editing interface, and the editing interface is scheduled and adjusted to form scheduling data with information flow node information.

2. The adaptive optimization method for programming interface layout based on machine learning according to claim 1, characterized in that, The scheduling data is used to train an interface layout prediction model using a reinforcement learning algorithm to obtain the optimal interface layout data for predicting the current user behavior and device state, including: Obtain user behavior features and device status parameters contained in the scheduling data, and encode the user behavior features and device status parameters into a dynamic state vector; The arrangement mode, spatial priority weight, and resource distribution strategy of the interface components in the current editing interface are used as a set of adjustable actions to make multi-dimensional layout adjustments to the current editing interface. Based on the task completion time in the task set T and the hotspot coverage in the current editing interface, a user experience quantitative index is constructed, and combined with the device status parameters, a system load balancing coefficient is calculated to construct an interface layout prediction model. The interface layout prediction model is trained using a near-end strategy optimization algorithm combined with historical interaction data. The convergence optimal strategy of the interface layout prediction model is calculated through a composite reward function to obtain the optimal interface layout data that constrains the current editing interface.

3. The adaptive optimization method for programming interface layout based on machine learning according to claim 2, characterized in that, The near-end strategy optimization algorithm assesses the importance of task information in the current editing interface by evaluating the coverage of click hotspots in the current editing interface, including: Based on the hotspot areas, click hotspot coverage, and user behavior characteristic indicators of the current editing interface, the importance of the task information of the current editing interface is assessed, and its expression is: ； in, This represents the number of clicks within the j-th hotspot area. This indicates the total number of clicks made in the current editing interface during the user's task. This indicates the importance of the j-th hotspot area based on user behavior characteristics.

4. The adaptive optimization method for programming interface layout based on machine learning according to claim 2, characterized in that, A particle swarm optimization algorithm based on global search is used to construct an interface layout prediction model to seek the optimal solution, including: The optimal interface layout data is initialized, an initial population is randomly generated, and the user experience quantification index is taken as a collective of m particles, with the particle number set. The interface layout prediction model is trained by combining historical interaction data. All particles are diffused at a rate v in the D-dimensional search space. The particle swarm optimization process is performed to update the velocity and position of each particle. A linearly decreasing weighting method is used to introduce inertia weights for global search of the particle swarm. The number of iterations of the particle swarm changes as follows: ； in, This represents the initial inertia weight value. This represents the inertia weight value after the particle swarm optimization algorithm has iterated to its maximum number of iterations. Indicates the maximum number of iterations. Indicates the current iteration number; By examining the historical optimal points of all particles, the optimal path is obtained, and the coverage of the click hotspot and the task completion time in the current editing interface are adjusted.

5. The adaptive optimization method for programming interface layout based on machine learning according to claim 4, characterized in that, The optimal interface layout data is adjusted into visual layout adjustment instructions, and the arrangement, size ratio, and hierarchical relationship of each functional module in the programming interface are dynamically updated, including: The optimal particle position vector output by the interface layout prediction model is parsed into specific interface parameter configurations, and formatted and encapsulated through a preset layout instruction protocol to generate visual layout script code that can be executed on the front end. A visual rendering scheduling module is integrated into the programming interface system. It receives layout adjustment instructions from the layout prediction model and maps them to the corresponding functional module rendering actions. Based on the position information in the layout adjustment instructions, the functional modules are rearranged in an instant switching manner to maintain the visual coherence of the overall interface and the continuity of user operation. Based on the user's current interaction context, the display level and focus of each module are set, and a user confirmation feedback channel is set. If the user manually adjusts or triggers a rollback operation, the current adjustment behavior is recorded as a training sample and synchronously fed back to the interface layout prediction model.

6. The adaptive optimization method for programming interface layout based on machine learning according to claim 5, characterized in that, By receiving user feedback, the interface layout prediction model is incrementally learned online to adaptively adjust the programmed interface layout, including: Real-time collection of user feedback behavior data on the current layout; cleaning and normalization of the collected feedback data; and combining it with the original input feature vector to form a complete feedback sample. The incremental random forest algorithm is used to add the feedback samples one by one into the training process, fine-tuning the key parameters of the interface layout prediction model and outputting a new optimal layout scheme. Based on the optimal layout scheme, a visual layout adjustment command is triggered to automatically optimize the position, size, and hierarchy of each functional module in the programming interface, thereby achieving continuous adaptive evolution of the interface layout.

7. A machine learning-based adaptive optimization system for programming interface layout, used to implement the machine learning-based adaptive optimization method for programming interface layout as described in any one of claims 1-6, characterized in that, include: The multi-source data acquisition module is used to collect multi-source interactive behavior data of users in the programming interface in real time; The feature extraction and vectorization module cleans, normalizes, and encodes the collected multi-source data, extracts user behavior features, interface status features, and personalized preference features, and constructs standardized feature vectors as input samples for machine learning models. The layout prediction modeling module constructs an interface layout prediction model based on the particle swarm optimization algorithm. It takes historical user behavior data and current feature vectors as input and searches for the optimal interface configuration scheme in a high-dimensional layout parameter space. The visualization instruction generation module parses the optimal layout data into visual layout adjustment instructions that can be executed on the front end, describes the interface components in JSON format, and passes them to the interface control layer through the rendering engine interface. The interface dynamic update module dynamically updates the arrangement, size ratio, and display hierarchy of each functional module in the programming interface based on the received layout adjustment instructions.

Images