Terminal equipment interface display method and system based on user habits

Abstract

The invention discloses a terminal equipment interface display method and system based on user habits, and belongs to the technical field of digital display, key features such as user intention deviation degree and interface attention stability index are extracted by collecting behavior data of a user for interaction objects and function modules, and a user preference weight table is calculated and dynamically updated; interface module sorting is further carried out in combination with a deep learning model, a fuzzy logic evaluation mechanism is introduced to carry out interpretable scoring on a sorting result, when a scoring result is lower than a threshold value, the system automatically corrects a sorting strategy based on user feedback, and finally optimized interface display content is output. Personalization, dynamics and controllability of interface display are achieved, and the man-machine interaction experience and the user satisfaction degree of the terminal equipment are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of digital display technology, and in particular to a terminal device interface display method and system based on user habits. Background Art

[0002] User-habit-based terminal device interface display refers to the ability of terminal devices (such as mobile phones, smart devices, and friend phones) to intelligently adjust the interface layout, function access, and content presentation based on the user's usage habits, preferences, and interactive behaviors, thereby providing a more personalized and convenient operating experience. Taking the friend phone as an example, the device can automatically prioritize the corresponding contact cards, quick reply buttons, or commonly used function modules on the main interface based on the user's frequently used contacts, applications, chat frequency, and other data, allowing users to interact with close friends more efficiently.

[0003] The existing technology has the following shortcomings:

[0004] In the existing technology, when sorting and optimizing the display of terminal device interface modules based on deep learning models (such as reinforcement learning or neural networks), there are problems such as unexplainable decision paths and opaque model adjustments, which can easily lead to abnormal interface display logic. For example, in terminals that emphasize intimate social interaction, such as girlfriends' phones, misjudgment of internal model weights may cause non-core modules (such as weather forecasts and advertising push) to be displayed first, while real high-frequency contacts or core interactive functions are sunk or hidden, seriously affecting the user experience. However, existing systems generally lack an explainability mechanism and are unable to track the basis for sorting decisions. At the same time, there is a lack of emergency fallback mechanisms or manual intervention methods. It is also difficult to accurately adjust model parameters through user interface feedback, resulting in difficult system debugging and uncontrollable risks. Summary of the Invention

[0005] The purpose of the present invention is to provide a terminal device interface display method and system based on user habits to solve the shortcomings of the background technology.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solution: a method for displaying a terminal device interface based on user habits, comprising:

[0007] Collect user behavior data, including the number of clicks, dwell time, and frequency of module use on each interactive object and functional module;

[0008] Extracting characteristic parameters for enhancing the interpretability of the interface display model from the collected data, the characteristic parameters including user intention deviation and interface attention stability index;

[0009] Based on the user behavior data and the characteristic parameters, calculate the user preference weight value of each interactive object and functional module, and construct a user preference weight table;

[0010] Introducing a time decay factor to dynamically update preference weights;

[0011] Based on the preference weight table and feature parameters, a deep learning model is used to sort the modules of the interface and output an initial display decision;

[0012] Performing an explainability scoring process on the initial display decision, evaluating the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtaining an explainability score value;

[0013] When the explainability score is lower than a preset threshold, the display sorting strategy is modified in real time based on user interface feedback data;

[0014] Output optimized terminal device interface display content based on the sorting results and correction mechanism.

[0015] Preferably, the collection of user behavior data includes: real-time recording of the user's click behavior on each functional module in the interface, the click behavior including the click location, timestamp and module identification information; when the user enters or leaves a functional module, recording the entry and exit timestamps to calculate the stay time; within the set time window, counting the user's call frequency for each module to form a module usage frequency index.

[0016] Preferably, the method for extracting the user intention deviation IDS is:

[0017] In the set time window, the user's click frequency and usage time for each functional module or interactive object are counted to form a historical preference vector: H = [h1,h2,...,h i ,...,h n ]; where h i represents the historical average preference score of the user for the i-th module, and n is the total number of functional modules;

[0018] For the user behavior in the current time window, the behavior indicators with the same dimensions as the previous step are counted to form the current behavior vector: C = [c1, c2, ..., c i ,...,c n ]; where c i Indicates the user's current behavior score for the i-th module;

[0019] The cosine similarity is used to reflect the similarity between the historical preference vector and the current behavior vector, and the degree of deviation is expressed by subtracting the similarity from 1. The expression is: Where IDS is the user intention deviation.

[0020] Preferably, the method for obtaining the interface attention stability index IAS is: using a set time window, collecting the residence time of each module on a daily basis to form a time series: T i =[t i,1 ,t i,2 ,...,t i,j ,...,t i,k ]; where t i,j represents the user's stay time in module i on day j; k is the total number of days; for time series T i Calculate the variance of residence time Indicates the degree of user attention fluctuation, the expression is: in, is the average residence time of module i, and the stability index IAS is calculated. An exponential transformation is used to map low variance to high stability. The calculation expression is: Among them, α is the adjustment coefficient, which controls the sensitivity of fluctuations to stability.

[0021] Preferably, the number of clicks C is collected i , residence time D i and frequency of use F i ; Build an initial preference score based on user behavior The expression is: Among them: α, β, γ are the weighting coefficients of each indicator respectively; the modified user preference weight value P i The calculation expression is:

[0022] Preferably, a time decay factor is introduced to dynamically update the preference weight, specifically including:

[0023] Set the time decay factor to: f(t) = e -λt ; Where: t is the number of days from the time when the behavior occurs to the current time; λ is the time decay coefficient; for module i, its historical behavior record set is {b i,1 ,b i,2 ,...,b i,z}, the basic preference score of each record is s i,j , timestamp is t i,j , the cumulative preference weight after decay for: Where: T is the current time point; z represents the number of historical behavior records of the user on module i.

[0024] Preferably, the explainability scoring process includes:

[0025] The user intention deviation and attention stability index are used as input items and the fuzzy logic reasoning model is introduced;

[0026] Perform fuzzy reasoning according to preset language rules;

[0027] The centroid method is used for defuzzification to obtain the interpretability score ES.

[0028] Preferably, if ES ≥ threshold, it means that the sorting logic is reasonable and the interface is displayed normally; if ES < threshold, the sorting intervention mechanism is triggered to correct, roll back or mark the sorting result as abnormal.

[0029] Preferably, the real-time correction of the sorting strategy based on user interface feedback data includes:

[0030] Collect user feedback on the current interface sorting, including skipping modules, manually adjusting the order, and thumbs-up / down operations;

[0031] Set negative feedback marks for frequently skipped high-ranking modules to lower their ranking priority;

[0032] Increase the weight of modules that are frequently clicked but initially ranked low;

[0033] And without changing the overall structure, local rearrangement is performed on medium and low weight modules.

[0034] The present invention also provides a terminal device interface display system based on user habits, including a behavior data collection unit, a feature extraction and analysis unit, a preference weight calculation unit, a time decay adjustment unit, a depth sorting decision unit, an interpretability evaluation unit, a sorting correction unit, and an optimized interface output unit;

[0035] Behavior data collection unit: collects user behavior data, including the number of clicks, dwell time, and frequency of module use on each interactive object and functional module;

[0036] Feature extraction and analysis unit: extracts feature parameters from the collected data to enhance the interpretability of the interface display model, including user intention deviation and interface attention stability index;

[0037] Preference weight calculation unit: based on the user behavior data and the characteristic parameters, calculates the user preference weight value of each interactive object and functional module, and constructs a user preference weight table;

[0038] Time decay adjustment unit: introduces a time decay factor to dynamically update the preference weight;

[0039] Deep sorting decision unit: based on the preference weight table and feature parameters, uses a deep learning model to sort the modules of the interface and outputs an initial display decision;

[0040] Explainability evaluation unit: performs explainability scoring processing on the initial display decision, evaluates the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtains an explainability score value;

[0041] Sorting correction unit: when the explainability score is lower than a preset threshold, the display sorting strategy is corrected in real time based on the user interface feedback data;

[0042] Optimized interface output unit: outputs optimized terminal device interface display content based on the sorting results and correction mechanism.

[0043] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0044] 1. This invention collects multidimensional user behavior data on terminal devices and introduces user intention deviation and interface attention stability index as enhanced feature parameters to construct an interface display optimization mechanism with interpretability and dynamic adaptability. Compared with traditional black-box sorting methods that rely solely on click frequency or deep modeling, this invention achieves a detailed understanding and modeling of user behavior at all stages before, during, and after sorting. It maintains the timeliness of preferences through a time decay factor, significantly improving the matching degree between interface display content and users' actual preferences.

[0045] 2. The present invention implements the explainability scoring of interface sorting results through a fuzzy logic reasoning system, and introduces a real-time sorting correction mechanism based on user feedback. When anomalies occur in the sorting logic, it can quickly intervene and optimize to ensure that the interface presentation is stable, continuous and reliable. This not only improves the user's personalized experience, but also enhances the controllability and maintainability of the terminal system. It is particularly suitable for smart terminal devices such as girlfriend phones that emphasize the quality of human-computer interaction and emotional adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0047] Figure 1 This is a mind map of the method of the present invention.

[0048] Figure 2 This is a mind map of the system units of the present invention. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] Example 1, please refer to Figure 1 As shown, the terminal device interface display method based on user habits described in this embodiment includes:

[0051] Collect user behavior data, including the number of clicks, dwell time, and frequency of module usage of each interactive object and functional module;

[0052] Extracting characteristic parameters for enhancing the interpretability of the interface display model from the collected data, the characteristic parameters including user intention deviation and interface attention stability index;

[0053] Based on the user behavior data and the characteristic parameters, calculate the user preference weight value of each interactive object and functional module, and construct a user preference weight table;

[0054] Introducing a time decay factor to dynamically update preference weights;

[0055] Based on the preference weight table and feature parameters, a deep learning model is used to sort the modules of the interface and output an initial display decision;

[0056] Performing an explainability scoring process on the initial display decision, evaluating the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtaining an explainability score value;

[0057] When the explainability score is lower than a preset threshold, the display sorting strategy is modified in real time based on user interface feedback data;

[0058] Output optimized terminal device interface display content based on the sorting results and correction mechanism.

[0059] In this invention, in order to realize intelligent dynamic display of the terminal device interface according to user habits, it is necessary to collect and record the user's behavior during the device operation process in a refined manner. Specifically, it includes the following three core behavior indicators: number of clicks, dwell time, and module usage frequency. The collection methods and functions are as follows:

[0060] The number of clicks refers to the total number of times a user has a clear touch event on a specific interface element, functional module, or interactive object (such as a contact avatar, function button, card component, etc.).

[0061] Collection method: Through the touch event monitoring module of the terminal system, each click operation of the user in the interface is captured in real time; each click event contains information such as the corresponding module ID, timestamp, click location (X, Y coordinates), etc.; for dynamic content (such as sliding modules), click statistics of sub-components in the sliding area are added.

[0062] Collection frequency: can be configured to be counted on a daily or hourly basis, and supports partitioned collection based on interface version and layout structure.

[0063] Application function: It is used to determine the user's willingness to actively trigger each functional module. It is the most direct interest and preference signal; it can be used as an explicit weight parameter in preference modeling.

[0064] Dwell time refers to the length of time a user stays and browses a certain functional module or page, including the time spent reading content, observing information, waiting for loading, etc.

[0065] Collection method: record the entry timestamp when the user enters a certain interface or module; record the exit timestamp when the user leaves, jumps or closes; the difference between the two is the stay time; combine the scrolling behavior to determine whether it is a valid stay (excluding hovering without operation).

[0066] Collection frequency: Periodic collection for each page visit; supports summary analysis of user dimensions and module dimensions.

[0067] Application function: Reflects the user's passive attention to the module, which helps to identify the blind spots of user interests and weak points in content appeal; combined with the number of clicks, it can determine abnormal preferences of "high-viewing low points" or "high points and low-viewing".

[0068] Module usage frequency refers to the frequency at which users actually call a function or interact with a module within a period of time, including clicks, slides, voice interactions, inputs, and other behaviors.

[0069] Collection method: Set a time window (such as day, week, month) to normalize the comprehensive usage behavior of each module; including active use (such as initiating a chat, clicking a function button) and passive triggering (such as the system automatically recommending a module to be browsed); the system layer records the module call log, associating the user ID, module ID and function type.

[0070] Collection frequency: periodic synchronization (such as daily aggregation), or real-time cache update in combination with machine learning models.

[0071] Application function: As the core variable for long-term modeling of user usage habits, it is used to evaluate the stable preference trend of users for modules and is the basic factor for calculating the user preference weight table.

[0072] From the collected data, feature parameters for enhancing the interpretability of the interface display model are extracted. The feature parameters include user intention deviation and interface attention stability index, specifically including:

[0073] User intention deviation is used to measure the degree of difference between the user's current behavior pattern and his or her historical stable behavior. The higher the value, the more likely the current behavior is to be an "abnormal attempt" or "temporary use" and should not be overly privileged.

[0074] The extraction method of user intention deviation IDS is:

[0075] In a set time window (e.g., the past 7 days), the user's click frequency and usage time for each functional module or interactive object are counted to form a historical preference vector: H = [h1, h2, ..., h i ,...,h n ]; where h i represents the user's historical average preference score for the i-th module (e.g., the normalized weighted value of click counts + dwell time), and n is the total number of functional modules;

[0076] For the user behavior in the current time window, the behavior indicators with the same dimensions as the previous step are counted to form the current behavior vector: C = [c1, c2, ..., c i ,...,c n ]; where c i Represents the user's current behavior score for the i-th module.

[0077] The cosine similarity is used to reflect the similarity between the historical preference vector and the current behavior vector, and the degree of deviation is expressed by subtracting the similarity from 1. The expression is: Where IDS is the user intention deviation. IDS∈[0,1] is the deviation score, where a higher value indicates a greater deviation, indicating that the current behavior does not conform to the long-term preference.

[0078] The interface attention stability index is used to measure the continuity and stability of users' attention to a certain module or function over multiple time periods. The higher the value, the more consistent the user's attention is, and the module can be recommended first.

[0079] The method for obtaining the interface attention stability index IAS is as follows: based on a set time window (e.g., the past 7 days), the residence time of each module is collected daily to form a time series: T i =[t i,1 ,t i,2 ,...,t i,j,...,t i,k ]; where t i,j represents the user's stay time in module i on day j; k is the total number of days; for time series T i Calculate the variance of residence time Indicates the degree of user attention fluctuation, the expression is: in, is the average residence time of module i, and the stability index IAS is calculated. An exponential transformation is used to map low variance to high stability. The calculation expression is: Where α is the adjustment coefficient, which controls the sensitivity of fluctuations to stability. IAS∈(0,1] is the module’s attention stability index, where the closer to 1, the more stable it is.

[0080] User preference weight values are used to quantify users' attention to and usage tendencies for various interactive objects (such as contacts) and functional modules (such as photo taking, voice, and chat) in the terminal interface, providing a calculable basis for interface sorting and display logic optimization.

[0081] Assume that there are W interactive objects or functional modules in the terminal, and collect user behavior indicators in the following three dimensions:

[0082] Clicks C i , residence time D i and frequency of use F i ; Normalize these three dimensions (for example, using maximum and minimum normalization) to obtain: The same equation applies to

[0083] Introduce the two feature parameters mentioned above to enhance interpretability:

[0084] User Intention Deviation IDS i (range: 0-1);

[0085] Focus on the stability index IAS i (Range: (0,1]);

[0086] These two parameters are used to dynamically adjust the credibility of users' short-term behaviors to prevent abnormal behaviors from affecting ranking.

[0087] Build an initial preference score based on user behavior The expression is: Among them: α, β, γ are the weighting coefficients of each indicator (such as 0.3, 0.3, 0.4); the coefficients can be dynamically adjusted to adapt to different user behavior patterns.

[0088] Corrected user preference weight value P i The calculation expression is: If a module has high attention stability (IAS close to 1) and low intention deviation (IDS close to 0), the final preference score will be maintained; if a module has high intention deviation (IDS close to 1) or unstable attention (IAS close to 0), the score will decrease;

[0089] Set the preference weights P of all modules or objects i Record according to module ID or label to form a user preference weight table.

[0090] The time decay factor is introduced to dynamically update the preference weight, including:

[0091] In the terminal device, the initial preference weight of each interactive object or functional module is calculated based on the user's behavioral data within a certain time window (such as the past 7 days).

[0092] The exponential decay model is used to make the weight of the behavior smaller the longer it is. The time decay factor is: f(t) = e -λt ; Where: t is the number of days from the time the behavior occurs to the current time; λ is the time decay coefficient, which controls the weight decay speed, and the recommended value is between 0.05 and 0.3; f(t)∈(0,1], the further the time, the smaller the factor.

[0093] In user behavior data, each record contains timestamp information. For module i, its historical behavior record set is {b i,1 ,b i,2 ,...,b i,z}, the basic preference score of each record is s i,j , timestamp is t i,j The cumulative preference weight after decay for: Where: T is the current time point (in days); z represents the number of historical behavior records of users on module i, that is, the number of times the user interacted with the module in the past period of time (such as the past 30 days).

[0094] All modules Perform normalization processing and construct a new user preference weight table.

[0095] The attenuation weight is recalculated each time a user behavior occurs or reaches a set update cycle (such as every morning). The system can set a maximum number of days for storage (such as 30 days), and data that exceeds the period will no longer be included in the calculation. It can be linked with the user intention deviation (IDS). If the recent behavior deviates significantly, the upper limit of its weight in the attenuation model will be reduced.

[0096] By integrating the user preference weight table with interpretable feature parameters, a deep learning sorting model is constructed to achieve intelligent sorting and display logic optimization of interface modules, and finally output the initial display decision for dynamic interface rendering.

[0097] For each interactive object or functional module M in the interface i , construct a multi-dimensional input feature vector x i , including the following:

[0098] Behavioral preference characteristics (obtained from the weight table): normalized click count, dwell time, and frequency of use; calculated initial user preference weight value P i ; User intention deviation; Interface attention stability index.

[0099] Regression ranking model based on fully connected neural network;

[0100] Ranking networks based on the Learning to Rank (LTR) concept (such as RankNet, LambdaRank, and DeepFM-Rank); attention mechanism models that enhance interpretability;

[0101] Model input: feature vector x of each module i ;

[0102] Model output: ranking score S for each module i ;

[0103] S i =f(x i ;θ); where f represents the neural network function and θ is the model parameter.

[0104] For all modules, the ranking score S i Sort in descending order;

[0105] Get sorted index sequence p is the total number of functional modules in the sorted list;

[0106] Sort by scores and place the top k modules on the main interface (such as the shortcut bar and main card area);

[0107] Medium and low-weight modules are sunk or moved to more areas;

[0108] Special weight modules (such as whitelist objects) can be inserted into designated locations for display fixation.

[0109] Perform interpretability scoring on the initial display decision. Based on the user intention deviation and the interface attention stability index, evaluate the rationality of the current interface sorting logic and obtain an interpretability score value, which specifically includes:

[0110] The user intention deviation IDS and the interface attention stability index IAS are used as the input items of fuzzy logic; the explainability score ES is used as the output item of fuzzy logic;

[0111] Map IDS and IAS to the corresponding linguistic variables in the fuzzy set, such as:

[0112] IDS is classified as: low (L), medium (M), and high (H);

[0113] IAS is divided into: low (L), medium (M), and high (H);

[0114] For example, IDS = 0.1 → high membership in the "low" set; IAS = 0.8 → high membership in the "high" set. Numerical values are mapped to membership degrees through membership functions (such as triangular or trapezoidal functions) for use in fuzzy reasoning.

[0115] Define several rules to represent the impact of the combination of IDS and IAS on ES. Example rules are as follows:

[0116] If IDS is low and IAS is high, ES is high;

[0117] If IDS is high and IAS is low, ES is low;

[0118] If IDS is medium and IAS is medium, then ES is medium;

[0119] If IDS is high and IAS is high, ES is medium to low;

[0120] If IDS is low and IAS is low, ES is medium to low.

[0121] Rules can be set by experience or expert systems and are scalable.

[0122] Substitute the fuzzified input items into the fuzzy rule system, calculate the activation strength of each rule (that is, the degree to which each rule satisfies the conditions), and use reasoning methods such as the Mamdani model for synthesis:

[0123] For each activated rule, calculate its fuzzy conclusion on the output item ES;

[0124] All conclusions are superimposed and synthesized in the output space to form a fuzzy output distribution.

[0125] The fuzzy output distribution is converted into the final interpretability score ES. The commonly used method is the centroid method, and the expression is: Among them, μ ES(x) is the membership degree of the ES fuzzy output function at value x. The score ES∈[0,1]. The higher the value, the more consistent the current interface ranking result is with the user's preferences and behavior patterns, and the more reasonable the ranking is.

[0126] If ES ≥ threshold (such as 0.6), it means that the sorting logic is reasonable and the interface is displayed normally; if ES < threshold, the sorting intervention mechanism is triggered to correct, roll back or mark the sorting results as abnormal.

[0127] When the explainability score is lower than a preset threshold, the display sorting strategy is modified in real time based on user interface feedback data.

[0128] The system actively or passively collects user feedback on the current interface ranking, including but not limited to: quickly swiping to skip a module; manually dragging to adjust the module order; frequently clicking on lower-level modules or ignoring top-level modules; and clicking on user subjective feedback buttons such as "Recommend Irrelevant" and "Show Dislike." These feedback behaviors are mapped into corrective signals, such as feedback reinforcement signals (indicating a ranking increase) and feedback suppression signals (indicating a ranking decrease).

[0129] Based on feedback signals, the current ranking strategy is modified in real time. Specifically, this includes temporarily increasing the preference weight of undervalued but highly clicked or highly feedbacked modules; and lowering the display priority of overvalued but ignored or skipped modules. The ranking strategy also preserves the position of modules with higher credibility scores; and only partially repositions modules in the middle or lower layers. This avoids global disruption of the ranking strategy to ensure user perception continuity.

[0130] Abnormal sorting is recorded in the user behavior log for subsequent model training and optimization. Differentiated strategies are implemented based on user tags, such as adopting more proactive correction strategies for “sensitive users.”

[0131] Based on the revised sorting strategy, the terminal device's main interface display content is regenerated; each functional component, interactive object card or recommended content area is rendered according to the new module sorting position; support is provided for instant rendering (such as front-end refresh), or the optimization results are automatically presented the next time the interface is opened; the system records the optimization effect and observes subsequent user feedback.

[0132] Example 2, please refer to Figure 2 As shown, the terminal device interface display system based on user habits described in this embodiment includes a behavior data collection unit, a feature extraction and analysis unit, a preference weight calculation unit, a time decay adjustment unit, a depth sorting decision unit, an interpretability evaluation unit, a sorting correction unit, and an optimized interface output unit;

[0133] Behavior data collection unit: collects user behavior data, including the number of clicks, dwell time, and frequency of module use on each interactive object and functional module;

[0134] Feature extraction and analysis unit: extracts feature parameters from the collected data to enhance the interpretability of the interface display model, including user intention deviation and interface attention stability index;

[0135] Preference weight calculation unit: based on the user behavior data and the characteristic parameters, calculates the user preference weight value of each interactive object and functional module, and constructs a user preference weight table;

[0136] Time decay adjustment unit: introduces a time decay factor to dynamically update the preference weight;

[0137] Deep sorting decision unit: based on the preference weight table and feature parameters, uses a deep learning model to sort the modules of the interface and outputs an initial display decision;

[0138] Explainability evaluation unit: performs explainability scoring processing on the initial display decision, evaluates the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtains an explainability score value;

[0139] Sorting correction unit: when the explainability score is lower than a preset threshold, the display sorting strategy is corrected in real time based on the user interface feedback data;

[0140] Optimized interface output unit: outputs optimized terminal device interface display content based on the sorting results and correction mechanism.

[0141] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0142] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0143] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0144] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for displaying a terminal device interface based on user habits, characterized in that: include: Collect user behavior data, including the number of clicks, dwell time, and frequency of module use on each interactive object and functional module; Extracting characteristic parameters for enhancing the interpretability of the interface display model from the collected data, the characteristic parameters including user intention deviation and interface attention stability index; Based on the user behavior data and the characteristic parameters, calculate the user preference weight value of each interactive object and functional module, and construct a user preference weight table; Introducing a time decay factor to dynamically update preference weights; Based on the preference weight table and feature parameters, a deep learning model is used to sort the modules of the interface and output an initial display decision; Performing an explainability scoring process on the initial display decision, evaluating the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtaining an explainability score value; When the explainability score is lower than a preset threshold, the display sorting strategy is modified in real time based on user interface feedback data; Output optimized terminal device interface display content based on the sorting results and correction mechanism.

2. The method for displaying a terminal device interface based on user habits according to claim 1, characterized in that: The collection of user behavior data includes: real-time recording of user click behavior on each functional module in the interface, wherein the click behavior includes click location, timestamp and module identification information; when the user enters or leaves a functional module, the entry and exit timestamps are recorded to calculate the residence time; within a set time window, the frequency of user calls to each module is counted to form a module usage frequency index.

3. The method for displaying a terminal device interface based on user habits according to claim 1, characterized in that: The extraction method of user intention deviation IDS is: In the set time window, the user's click frequency and usage time for each functional module or interactive object are counted to form a historical preference vector: H = [h1,h2,...,h i ,...,h n ]; where h i represents the historical average preference score of the user for the i-th module, and n is the total number of functional modules; For the user behavior in the current time window, the behavior indicators with the same dimensions as the previous step are counted to form the current behavior vector: C = [c1, c2, ..., c i ,...,c n ]; where c i Indicates the user's current behavior score for the i-th module; The cosine similarity is used to reflect the similarity between the historical preference vector and the current behavior vector, and the degree of deviation is expressed by subtracting the similarity from 1. The expression is: Where IDS is the user intention deviation.

4. The method for displaying a terminal device interface based on user habits according to claim 3, characterized in that: The method for obtaining the interface attention stability index IAS is as follows: based on a set time window, the residence time of each module is collected daily to form a time series: T i =[t i,1 ,t i,2 ,...,t i,j ,...,t i,k ]; where t i,j represents the user's stay time in module i on day j; k is the total number of days; for time series T i Calculate the variance of residence time Indicates the degree of user attention fluctuation, the expression is: in, is the average residence time of module i, and the stability index IAS is calculated. An exponential transformation is used to map low variance to high stability. The calculation expression is: Among them, α is the adjustment coefficient, which controls the sensitivity of fluctuations to stability.

5. The method for displaying a terminal device interface based on user habits according to claim 4, characterized in that: Collect click count C i , residence time D i and frequency of use F i ; Build an initial preference score based on user behavior The expression is: Among them: α, β, γ are the weighting coefficients of each indicator respectively; the modified user preference weight value P i The calculation expression is:

6. The method for displaying a terminal device interface based on user habits according to claim 5, characterized in that: The time decay factor is introduced to dynamically update the preference weight, including: Set the time decay factor to: f(t) = e -λt ; Where: t is the number of days from the time when the behavior occurs to the current time; λ is the time decay coefficient; for module i, its historical behavior record set is {b i,1 ,b i,2 ,...,b i,z }, the basic preference score of each record is s i,j , timestamp is t i,j , the cumulative preference weight after decay for: Where: T is the current time point; z represents the number of historical behavior records of the user on module i.

7. The method for displaying a terminal device interface based on user habits according to claim 6, characterized in that: The explainability scoring process includes: The user intention deviation and attention stability index are used as input items and the fuzzy logic reasoning model is introduced; Perform fuzzy reasoning according to preset language rules; The centroid method is used for defuzzification to obtain the interpretability score ES.

8. The method for displaying a terminal device interface based on user habits according to claim 7, characterized in that: If ES ≥ the threshold, it means that the sorting logic is reasonable and the interface is displayed normally; If ES is less than the threshold, the sorting intervention mechanism is triggered to correct, roll back or mark the sorting results as abnormal.

9. The method for displaying a terminal device interface based on user habits according to claim 8, characterized in that: The real-time correction of the sorting strategy based on user interface feedback data includes: Collect user feedback on the current interface sorting, including skipping modules, manually adjusting the order, and thumbs-up / down operations; Set negative feedback marks for frequently skipped high-ranking modules to lower their ranking priority; Increase the weight of modules that are frequently clicked but initially ranked low; And without changing the overall structure, local rearrangement is performed on medium and low weight modules.

10. A terminal device interface display system based on user habits, used to implement the terminal device interface display method based on user habits according to any one of claims 1 to 9, characterized in that: It includes a behavior data collection unit, a feature extraction and analysis unit, a preference weight calculation unit, a time decay adjustment unit, a depth sorting decision unit, an explainability evaluation unit, a sorting correction unit, and an optimization interface output unit; Behavior data collection unit: collects user behavior data, including the number of clicks, dwell time, and frequency of module use on each interactive object and functional module; Feature extraction and analysis unit: extracts feature parameters from the collected data to enhance the interpretability of the interface display model, including user intention deviation and interface attention stability index; Preference weight calculation unit: based on the user behavior data and the characteristic parameters, calculates the user preference weight value of each interactive object and functional module, and constructs a user preference weight table; Time decay adjustment unit: introduces a time decay factor to dynamically update the preference weight; Deep sorting decision unit: based on the preference weight table and feature parameters, uses a deep learning model to sort the modules of the interface and outputs an initial display decision; Explainability evaluation unit: performs explainability scoring processing on the initial display decision, evaluates the rationality of the current interface sorting logic based on the user intention deviation and the interface attention stability index, and obtains an explainability score value; Sorting correction unit: when the explainability score is lower than a preset threshold, the display sorting strategy is corrected in real time based on the user interface feedback data; Optimized interface output unit: outputs optimized terminal device interface display content based on the sorting results and correction mechanism.

Images