Preloading method and device, electronic equipment and computer readable storage medium

Abstract

Embodiments of the present application disclose an application preloading method and device, electronic equipment and computer readable storage medium, and relate to the technical field of computers; the method comprises: obtaining historical behavior data of a current user, and determining a behavior score of each candidate application based on the historical behavior data; obtaining current environment perception data, and determining a correction coefficient of each candidate application based on the current environment perception data and information of a plurality of candidate applications; correcting the behavior score of each candidate application based on the correction coefficient to obtain a final score of each candidate application; determining a target application based on the final score, and preloading the target application. In this way, the present application can improve the hit rate of preloaded applications.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to an application preloading method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] The app launch speed on smart TVs is a key factor affecting user experience. Related technologies follow the pre-loading logic of smartphones, typically predicting potentially used apps and loading them into memory before user interaction to shorten launch time. However, unlike smartphones, which are "personal devices," televisions are "household devices," thus the pre-loading logic based on smartphones suffers from a low pre-loading hit rate. Summary of the Invention

[0003] This application provides an application preloading method, apparatus, electronic device, and computer-readable storage medium, which can improve the hit rate of preloaded applications.

[0004] In a first aspect, embodiments of this application provide an application preloading method, including: Obtain the current user's historical behavior data, and determine the behavior score of each candidate application based on the historical behavior data; Acquire current environmental perception data, and based on the current environmental perception data and information of multiple candidate applications, determine the correction coefficient for each candidate application; The behavioral scores of each candidate application are corrected based on the correction coefficient to obtain the final scores of each candidate application. The target application is determined based on the final score, and the target application is preloaded.

[0005] Secondly, embodiments of this application provide an application preloading device, comprising: The acquisition module is used to acquire the current user's historical behavior data and determine the behavior score of each candidate application based on the historical behavior data. The determination module is used to acquire current environmental perception data and, based on the current environmental perception data and information of multiple candidate applications, determine the correction coefficient of each candidate application. The correction module is used to correct the behavior scores of each candidate application based on the correction coefficient to obtain the final scores of each candidate application. A preloading module is used to determine the target application based on the final score and preload the target application.

[0006] Thirdly, embodiments of this application also provide an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in the application preloading method described above.

[0007] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the application preloading method described above.

[0008] Fifthly, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in embodiments of this application.

[0009] The embodiments of this application have the following beneficial effects: By acquiring historical behavioral data of the current user to determine the behavioral scores of each candidate application, an accurate user-side data foundation can be provided for preloading decisions based on the actual usage of the current user rather than all users, ensuring that the target application for preloading matches the current user's preferences. Acquiring current environment awareness data and combining it with information from candidate applications to determine correction coefficients enables the system to perceive the current usage scenario and capture real-time intent information that cannot be obtained from historical data alone. The behavioral scores are then corrected based on the correction coefficients to obtain the final score, achieving an effective fusion of long-term user preferences and real-time environment awareness. This ensures that the final score reflects both user usage patterns and adapts to the current usage scenario, significantly improving the accuracy of preloading predictions. Target applications are determined and preloaded based on the final score, ensuring that the application that best meets the user's current needs is preloaded, improving application startup speed while avoiding the waste of memory resources and system performance degradation caused by blind preloading. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram illustrating the steps of an application preloading method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating an application preloading method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an application preloading device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0013] In one embodiment, such as Figure 1 As shown, an application preloading method is provided. Although the logical order is illustrated in the step diagram, in some cases, the steps shown or described may be performed in a different order than that shown in the figures. Specifically, this application preloading method can be applied to smart terminal devices, which may include, but are not limited to, one or more of smart TVs, desktop computers, and in-vehicle computers. These will be described in detail below. It should be noted that the order of description of the following embodiments is not intended to limit the priority of the embodiments.

[0014] according to Figure 1 The application preloading method shown includes at least steps S110 to S140, which are described in detail below: In step S110, the historical behavior data of the current user is obtained, and the behavior score of each candidate application is determined based on the historical behavior data.

[0015] In step S120, current environmental perception data is acquired, and correction coefficients for each of the candidate applications are determined based on the current environmental perception data and information of multiple candidate applications.

[0016] In step S130, the behavioral scores of each candidate application are corrected based on the correction coefficient to obtain the final scores of each candidate application.

[0017] In step S140, the target application is determined based on the final score, and the target application is preloaded.

[0018] The system can identify the current user's identity and then filter the user's historical behavior data from application usage logs. In one embodiment, this could involve identifying the current user's facial features and then determining the user's identity based on those features. In another embodiment, it could involve obtaining the currently logged-in user's information and determining the user's identity based on that information.

[0019] Historical behavior data records the behavior of one or more users on smart terminal devices when using various applications. This may include, but is not limited to, one or more of the following: application launch count, duration of each use, usage time distribution, and application exit methods. Candidate applications refer to applications already installed on the smart terminal device and meeting the conditions for pre-loading. The behavior score is a quantitative value calculated based on historical behavior data, reflecting the user's preference for each candidate application; a higher score indicates a stronger user inclination to use that application.

[0020] Specifically, the system can extract historical behavior data of the current user from the application usage logs of smart terminal devices. For each candidate application, it can calculate indicators such as the frequency of use and cumulative usage time, and convert the statistical indicators into behavior scores based on preset scoring rules. Optionally, behavior scores can be calculated for all installed applications one by one, or only for a subset of recently used applications.

[0021] Current environmental perception data refers to data acquired in real-time by environmental acquisition devices on smart terminal devices, used to characterize the current usage environment; examples include environmental images captured by cameras, environmental sounds captured by microphones, and environmental parameters acquired by various sensors. The correction coefficient is a weight adjustment factor output to each candidate application after inferring the user's real-time intent based on the current environmental perception data. It is used to make contextual adjustments based on the behavior score. The closer a candidate application matches the user's intent, the larger the corresponding correction coefficient, resulting in a higher final score for that candidate application.

[0022] The system can input current environmental perception data and information from multiple candidate applications (such as application lists, application category tags, etc.) into a semantic analysis model. The semantic analysis model will then perform scene semantic recognition on the environmental perception data, understand the user's intent, and output the corresponding correction coefficient based on the degree of matching between each candidate application and the user's intent.

[0023] The correction coefficients for each candidate application are calculated using a pre-defined mathematical relationship with their corresponding behavioral scores to obtain the final score for each application. This final score is a comprehensive score adjusted by the correction coefficients, integrating the user's historical behavioral preferences and current environmental awareness. Compared to a single behavioral score, it more accurately reflects the user's actual need for each candidate application in the current scenario. For example, the correction coefficient is multiplied by the behavioral score; a correction coefficient greater than 1 positively increases the behavioral score, while a correction coefficient less than 1 negatively decreases it.

[0024] Candidate applications are ranked from highest to lowest based on their final scores. At least one candidate application with the highest score is selected as the target application, and preloading is performed on it. This allows the system to respond quickly based on preloaded resources when the user subsequently initiates a launch command for the target application, significantly shortening the application's launch time from click to availability. The target application is selected from the candidate applications based on the final score and is deemed the most likely application for the current user. Preloading refers to loading some or all of the target application's process resources into the device's RAM before the user actively initiates the application launch operation.

[0025] In one embodiment, the smart terminal device is a smart TV equipped with a camera to capture images of the living room environment. The smart TV comes pre-installed with multiple applications, including live sports, online education, video playback, short videos, children's animation, and home fitness. When a user turns on the smart TV and enters the main interface, the system automatically triggers an application pre-loading process. First, the system captures the living room scene through the camera, performs face detection and extracts facial features, matches these features with those of registered users, and confirms that the current user is a registered user. Then, the system obtains the user's application usage logs for the past 30 days as historical behavior data. Based on this historical behavior data, a time decay model is used to calculate the behavior score for each candidate application. Each usage behavior is assigned a base score (e.g., 1 point for launching the application, 5 points for using it for more than 30 minutes, and -2 points for exiting immediately). An exponential decay weighting is applied based on the number of days since each usage behavior, giving higher weight to recent behaviors. For example, the behavior scores for each application are as follows: Live Sports 0.85, Children's Animation 0.72, Video Playback 0.65, Short Videos 0.40, and Online Education 0.30.

[0026] Next, the current environmental perception data is acquired. The current living room scene captured by the camera and the list of candidate applications are input into the semantic analysis model deployed on the device. The semantic analysis model identifies the scene semantic information in the scene: an adult male is sitting on the sofa, holding a remote control, with a relaxed expression, and the living room is dimly lit. Based on this, the semantic analysis model infers that the user's current intention is to quietly watch video content and outputs correction coefficients for each candidate application: sports live broadcast 0.8 (holding a remote control and in a quiet state, not a perfect match for an intense game scene), movie playback 1.2 (dim lighting and a relaxed expression are suitable for watching movies), children's animation 0.6 (the current scene is not dominated by children), short videos 0.5 (the user's posture indicates a long period of immersive viewing rather than short-term scrolling), and online education 0.3 (a relaxed state is not suitable for learning).

[0027] Multiplying each correction coefficient by its corresponding behavioral score yields the final score for each candidate application: Sports Live Streaming 0.85 × 0.8 = 0.68, Children's Animation 0.72 × 0.6 = 0.43, Movie Playback 0.65 × 1.2 = 0.78, Short Videos 0.40 × 0.5 = 0.20, and Online Education 0.30 × 0.3 = 0.09. After sorting by the final score, Movie Playback, with the highest score of 0.78, was selected as the target application. Preloading the Movie Playback application allows the system to quickly load the user interface based on the preloaded application when the user subsequently clicks its icon, significantly reducing startup time.

[0028] The technical solution adopted in this application determines the behavioral scores of each candidate application by acquiring the historical behavioral data of the current user. This provides an accurate user-side data foundation for preloading decisions based on the actual usage of the current user rather than all users, ensuring that the target application for preloading matches the current user's preferences. Acquiring current environment awareness data and combining it with candidate application information to determine correction coefficients enables the system to perceive the current usage scenario and capture real-time intent information that cannot be obtained from historical data alone. The behavioral scores are corrected based on the correction coefficients to obtain the final score, effectively integrating long-term user preferences with real-time environment awareness. This ensures that the final score reflects both user usage patterns and adapts to the current usage scenario, significantly improving the accuracy of preloading prediction. Determining the target application based on the final score and preloading it ensures that the application most suitable for the user's current needs is preloaded, improving application startup speed while avoiding memory resource waste and system performance degradation caused by blind preloading.

[0029] Based on the above technical solution, as an embodiment, determining the behavior score of each candidate application based on the historical behavior data may include: assigning a basic score to each historical usage behavior of each candidate application based on a time decay model; determining the weight of each basic score according to the interval between each historical usage behavior of each candidate application and the current time; wherein the weight of the historical usage behavior that is closer to the current time is greater; and performing a weighted summation of the basic scores of each historical usage behavior of each candidate application based on the weights to obtain the behavior score of each candidate application.

[0030] The time decay model is a mathematical model that dynamically assigns weights based on the interval between the time of the action and the current time. Its core idea is that actions that are closer to the current time have a greater impact on the score.

[0031] Each candidate application can be assigned a base score based on its historical usage behavior. This base score is a quantitative assessment of a single usage action, used to differentiate the positive and negative contributions of different types of usage behavior to application preference judgments. The base score can be determined based on the type of usage behavior. Specifically, the system extracts all historical usage records of each candidate application from the current user's application usage logs and assigns a base score to each usage action. Base scores can be differentiated based on the type of usage behavior. For example, launching an application assigns a positive score (e.g., 1 point), usage time exceeding a preset threshold assigns a higher positive score (e.g., 5 points for usage exceeding 30 minutes), and exiting immediately after launch assigns a negative score (e.g., -2 points). This differentiated assignment ensures that high-value usage behaviors receive higher base scores, while low-value or negative usage behaviors are identified and downweighted.

[0032] The weights of each base score are determined based on the interval between each historical usage action of each candidate application and the current time, with more recent historical usage actions corresponding to higher weights. These weights reflect the relative importance of each usage action in the final score; more recent usage actions have higher weights and a more significant impact on the final score. The number of days between each usage action and the current time can be calculated, and the corresponding weight can be calculated using an exponential decay function. The shorter the number of days (i.e., the more recent the action), the larger the output value of the exponential decay function, and the higher the weight; the larger the number of days (i.e., the more distant the action), the smaller the output value, and the lower the weight. The decay rate is controlled by a preset decay parameter; the larger the value of the decay parameter, the faster the weight of distant actions decays, meaning that the user's recent actions have a stronger dominance in the score.

[0033] The behavioral score for each candidate application is obtained by weighting and summing the base scores of each historical usage behavior based on weights. This approach balances the completeness and timeliness of user history. On the one hand, all historical usage behaviors are included in the calculation, avoiding insufficient sample size due to relying solely on recent behaviors. On the other hand, the time decay weighting highlights the dominance of recent behaviors, enabling the behavioral score to reflect the dynamic changes in user application preferences in a timely manner.

[0034] In one embodiment, assuming the current user is a registered "parent" user on a smart TV, and one of the candidate applications is a sports live streaming application, the system obtains all the user's historical usage records of the sports live streaming application in the past 30 days and assigns a base score to each usage behavior: 1st time (15 days ago): Launch the application, base score 1 point; 2nd time (14 days ago): Use for more than 30 minutes, base score 5 points; 3rd time (13 days ago): Exit immediately, base score -2 points; 4th time (10 days ago): Use for more than 30 minutes, base score 5 points; 5th time (5 days ago): Launch the application, base score 1 point; 6th time (2 days ago): Use for more than 30 minutes, base score 5 points.

[0035] The weights are calculated based on the number of days since each usage action, using an exponential decay function w=exp(-λ×d), where the decay parameter λ is a manually set value, which can be 0.1, and d is the number of days since the action: 1st time (d=15): w≈0.22; 2nd time (d=14): w≈0.25; 3rd time (d=13): w≈0.27; 4th time (d=10): w≈0.37; 5th time (d=5): w≈0.61; 6th time (d=2): w≈0.82.

[0036] The base score for each use behavior is multiplied by the corresponding weight and then summed to obtain the behavior score of the sports live streaming application: 1×0.22+5×0.25+(-2)×0.27+5×0.37+1×0.61+5×0.82=0.22+1.25-0.54+1.85+0.61+4.10=7.49.

[0037] The same calculation process was performed on the remaining candidate applications. Assume the final behavioral scores for each application are as follows: Sports Live Streaming 7.49, Children's Animation 6.20, Movie Playback 5.10, Short Videos 3.80, and Online Education 2.50. Sports Live Streaming received the highest behavioral score after time decay weighting due to its high recent usage frequency and numerous long-term usage behaviors, reflecting the user's current strong preference for the Sports Live Streaming application.

[0038] The technical solution adopted in this application uses a time decay model to assign decreasing weights to historical usage behaviors over time, giving recent behaviors a greater weight and accurately reflecting the dynamic changes in user application preferences. Each usage behavior is assigned a base score, distinguishing between different behavior types such as startup, prolonged use, and immediate exit, quantifying the actual contribution of each behavior to application preferences. The weights determined based on time intervals ensure that behaviors closer to the current time have a greater influence in the weighted summation, resulting in a behavior score that balances the completeness of historical behavior with the dominance of recent behavior, accurately reflecting the user's current application usage tendencies.

[0039] Based on the above technical solution, as an embodiment, obtaining the historical behavior data of the current user may include: collecting the facial data of the current user, extracting the facial features of the current user based on the facial data; matching the facial features with the facial features of registered users to determine the identity of the current user; and obtaining the historical behavior data corresponding to the identity.

[0040] The system captures an image of the current user using a camera on a smart terminal device. Facial data refers to image or video frame data containing the user's facial region, acquired through image acquisition devices such as cameras. Face detection is performed on this image to locate the facial region. Then, a pre-trained feature extraction model (such as a face recognition model based on a deep convolutional neural network) extracts facial features from the facial data. Facial features are a set of numerical vectors obtained after feature extraction from the facial data, uniquely representing the user's facial biometric characteristics, such as facial contours, facial proportions, and texture features.

[0041] The system stores the facial features of all registered users (registered users and their corresponding facial features are stored). Registered users are those who have pre-registered their facial features in the system. The system calculates the similarity between the current user's facial features and the facial feature templates of registered users. If the similarity exceeds a preset matching threshold, the current user is determined to be a registered user, and their corresponding identity identifier is obtained. This identity identifier is used to uniquely distinguish different users, and each identifier is associated with the user's independent historical behavioral data. If the similarity between the facial features of all registered users and the current user's facial features does not exceed the preset matching threshold, the current user is determined to be an unregistered user.

[0042] Based on the identified identity, search the user behavior database for the historical usage records associated with that identity, and extract the user's application usage logs within a preset time window as historical behavior data for subsequent behavior scoring calculations.

[0043] In one embodiment, three family members are registered on the smart TV: father (identifier User_A), mother (User_B), and child (User_C). Upon registration, the system captures facial images of each user via camera and extracts and stores corresponding facial feature templates. One evening, the father turns on the smart TV, and the TV camera automatically captures the living room scene. The system performs face detection on the image, locates the facial region, and extracts the current user's facial feature V_current. The system calculates the cosine similarity of V_current with the facial feature templates of User_A, User_B, and User_C: the similarity with User_A is 0.95, with User_B is 0.32, and with User_C is 0.28. Since the similarity with User_A (0.95) exceeds the preset matching threshold of 0.80, the system determines that the current user is the father and identifies him as User_A. Subsequently, the system retrieves User_A's application usage logs for the past 30 days from the user behavior database, including the number of times each candidate application is launched, usage duration, usage time period, and exit method, as historical behavior data for subsequent behavior scoring calculations. As a result, the system can accurately distinguish the specific family members currently using the TV, providing an accurate user identity basis for subsequent personalized preloading.

[0044] By adopting the technical solution of this application embodiment, by collecting facial data and extracting facial features for matching, the specific user identity of the currently used device can be automatically identified without the user having to manually switch accounts or log in; based on the identity identifier, the corresponding historical behavior data is obtained, enabling the system to maintain an independent behavior profile for each family member, and to achieve accurate personalized preloading decisions in the scenario of multiple users sharing devices in the family, significantly improving the preloading hit rate.

[0045] Based on the above technical solution, as an example, when the facial feature vector does not match the facial feature vector of a registered user, the age feature of the current user is identified; and the behavior scores of each candidate application corresponding to the age feature are searched.

[0046] "Facial features not matched with registered users" means that after calculating the similarity between the current user's facial features and the facial feature templates of all registered users, all calculation results do not exceed the preset matching threshold, meaning it is impossible to confirm that the current user is any registered user.

[0047] When the system determines that the current user's facial features do not match any registered users, it performs age recognition analysis on the collected facial data to obtain the current user's age characteristics. Age recognition analysis can be achieved through a pre-trained age classification model. This model, based on facial age-related features such as facial texture, facial contour, and skin features, classifies users into preset age groups. Age characteristics are information inferred from facial images about the user's age group, such as child, teenager, adult, or elderly, providing a coarse-grained user profile for pre-loading decisions in the absence of precise identity identifiers.

[0048] The system can maintain a general preference knowledge base, which stores behavioral scores for each candidate application corresponding to each age group. These behavioral scores are pre-set based on usage statistics of users in that age group or industry-wide general preference data. Based on the identified age characteristics, the system retrieves the behavioral scores of each candidate application corresponding to that age group from the general preference knowledge base and uses these scores as the current user's behavioral score for subsequent correction coefficient adjustments and pre-loading decisions.

[0049] Continuing with the previous example, a visitor (an unregistered user) turns on the smart TV. The TV camera automatically captures the living room scene and extracts the facial feature V_guest. The system calculates the cosine similarity between V_guest and the facial feature templates of three registered users (father User_A, mother User_B, and child User_C). The results are 0.45, 0.52, and 0.38, respectively, all below the preset matching threshold of 0.80. The system determines that the current user is unregistered. Age recognition analysis is then performed on the facial data corresponding to V_guest. The age classification model, based on facial texture and contour features, determines that the user belongs to the elderly age group. The system then queries the general preference knowledge base for the behavioral scores of various candidate applications corresponding to the elderly age group: opera playback 0.90, news broadcast 0.85, movie playback 0.70, health and wellness 0.65, sports live broadcast 0.40, short videos 0.35, online education 0.20, and children's animation 0.10. The system uses these scores as the current user's behavioral score for further processing. Therefore, even when faced with unfamiliar users who have never registered, the system can still provide targeted pre-loaded services based on age characteristics, rather than pre-loading applications completely randomly or indiscriminately.

[0050] By adopting the technical solution of this application embodiment, when the facial features do not match a registered user, the cold start preloading of new users is realized by identifying age features and searching for corresponding behavioral scores from the general preference knowledge base. Thus, even with zero historical data, targeted preloading services can still be provided based on age group preferences, filling the gap of traditional solutions that cannot effectively preload when there is no user data, and ensuring the user experience of all users.

[0051] Based on the above technical solution, as an embodiment, when the current user is a user group, the application preloading method may further include: obtaining the historical usage records of each of the candidate applications for the user group information; and determining the scenario coefficient based on the historical usage records.

[0052] The step of weighted summation of the basic scores of each candidate application's historical usage behavior based on the weights to obtain the behavior score of each candidate application may include: weighted summation of the basic scores of each candidate application's historical usage behavior based on the weights to obtain the basic behavior score of each candidate application; and weighted processing of the basic behavior scores of each candidate application based on the scenario coefficients to obtain the behavior score of each candidate application.

[0053] When the current user is a user group, scenario-weighted processing can be triggered. A user group refers to two or more users who are simultaneously in front of a smart terminal device and using the device together within the same time period, such as scenarios like father and son watching on the same screen, couples watching together, or the whole family gathering.

[0054] The system captures the current environment using a camera and performs face detection. When two or more faces are detected, it extracts facial feature vectors from each face and matches them against registered user identities to determine which registered users constitute the current user group. This user group information, representing the identities of currently present users, is determined by detecting and extracting features from multiple faces in the image using the camera and matching them one by one with the facial feature vector templates of registered users.

[0055] Based on the determined user combination information (i.e., the identities of each user currently present), the system searches the user behavior database for records of the user combination's joint use of various candidate applications within a historical period. For example, when the system detects that the current user combination is a parent-child combination, it searches for records of the frequency, duration, and usage methods of each candidate application when the parent-child combination was present in the past.

[0056] Based on the frequency and duration of shared use of each candidate application by the user group, a scenario coefficient is calculated for each candidate application. The scenario coefficient is a weighting adjustment factor determined based on the historical shared usage behavior of a specific user group, used to reflect the user group's preference for various candidate applications in shared usage scenarios. For example, if a parent-child group has a history of high shared use of game and animation applications, then game and animation applications are assigned a higher scenario coefficient (e.g., 1.3), while applications based on personal preferences are assigned a lower scenario coefficient (e.g., 0.8), and other applications are assigned a default scenario coefficient (e.g., 1.0).

[0057] Referring to the time-decay weighted summation method described above, the basic scores of each candidate application's historical usage behavior are multiplied by their corresponding time-decay weights and then summed to obtain the basic behavior score. Based on these weights, the basic scores of each candidate application's historical usage behavior are weighted and summed to obtain the basic behavior score for each candidate application. The basic behavior score is an intermediate score that has only undergone time-decay weighting but has not yet been adjusted for scenario coefficients; it reflects the degree of user preference for each candidate application at the individual user level.

[0058] The basic behavioral scores of each candidate application are weighted based on a scenario coefficient to obtain the behavioral score for each candidate application. For example, the basic behavioral score is multiplied by the scenario coefficient to obtain the final behavioral score after scenario adjustment. A scenario coefficient greater than 1 positively improves the basic behavioral score, a scenario coefficient less than 1 negatively suppresses it, and a scenario coefficient equal to 1 keeps the basic behavioral score unchanged.

[0059] In one embodiment, the father and child sit simultaneously in front of a smart TV. The camera detects both faces, extracts facial feature vectors, and matches them to determine the current user combination as "Father (User_A) + Child (User_C)". The system searches the user behavior database for shared usage records of "User_A + User_C" over the past 60 days. It finds that this combination uses children's animation apps 3 times per week, family fitness apps 2 times per week, and game apps 4 times per week, while sports live streaming apps and online education apps have almost no shared usage records under this combination. Based on this, the system determines the scenario coefficients for each candidate app: children's animation 1.3, family fitness 1.2, game apps 1.4, sports live streaming 0.7, online education 0.6, and video playback 1.0 (default). Then, following the time-decay weighted summation method of claim 2, the system calculates the basic behavior score for each candidate app based on the father User_A's historical usage data: sports live streaming 7.49, children's animation 4.20, video playback 5.10, game apps 3.50, online education 6.80, and family fitness 2.00.

[0060] Multiplying the basic behavior score by the corresponding scenario coefficient yields the scenario-weighted behavior score: Sports Live Streaming 7.49 × 0.7 = 5.24, Children's Animation 4.20 × 1.3 = 5.46, Movie Playback 5.10 × 1.0 = 5.10, Game Applications 3.50 × 1.4 = 4.90, Online Education 6.80 × 0.6 = 4.08, and Family Fitness 2.00 × 1.2 = 2.40. It can be seen that Children's Animation jumped from fourth to second place due to the increased weighting of the scenario coefficient of 1.3; Game Applications also saw a significant increase due to the scenario coefficient of 1.4; while Sports Live Streaming, although having the highest basic score at the father's individual level, ultimately dropped in ranking because the scenario coefficient for father-son co-screening was only 0.7. This scenario-weighted mechanism allows pre-loading decisions to adapt to the actual needs of multiple users using the app simultaneously.

[0061] By adopting the technical solution of this application embodiment, when multiple users are detected to be present at the same time, the scenario coefficient is determined by obtaining the common usage records of the user combination, and the basic behavior score is adjusted by scenario-based weighting. This expands the preloading decision from single individual preference to group scenario adaptation, thereby effectively solving the problem of decreased preloading hit rate in multi-user shared scenarios such as family gatherings, and improving the preloading priority of family-friendly applications.

[0062] Based on the above technical solution, as an embodiment, determining the correction coefficient of each candidate application based on the current environment perception data and the information of multiple candidate applications may include: inputting the current environment perception data and the information of multiple candidate applications into a semantic analysis model; identifying scene semantic information of the current environment perception data through the semantic analysis model; inferring user intent based on the scene semantic information; and determining the correction coefficient of each candidate application according to the user intent.

[0063] The semantic analysis model can input current environmental perception data (such as environmental images captured by cameras) and candidate application information together, allowing the model to perform inference. The semantic analysis model is a large-scale visual language model deployed on the edge of smart terminal devices, possessing deep semantic understanding capabilities. It can perform scene semantic recognition on environmental perception data and output structured analysis results. Candidate application information refers to a list of candidate applications and descriptive information such as category labels, used to assist the semantic analysis model in associating scene semantics with application types.

[0064] The semantic analysis model identifies the scene semantic information of the current environment perception data. The scene semantic information is a structured description output by the semantic analysis model after performing in-depth analysis of the environment perception data, and is used to characterize the fine-grained state of the current usage environment.

[0065] Scene semantic information may include, but is not limited to, one or more of the following: human emotions, held objects, and ambient lighting. Human emotions refer to the user's current facial expression or emotional state as identified by the model, such as relaxed, excited, or focused. Held objects refer to items held by the user as identified by the model, such as remote controls, game controllers, books, or mobile phones. Ambient lighting refers to the lighting conditions of the current environment as identified by the model, such as bright, dim, or low light. In practice, the semantic analysis model performs multi-dimensional analysis on the input environmental image, inferring emotional states from facial expressions, identifying the type of items held by the user in the image, determining ambient lighting conditions from the overall image brightness, and outputting analysis results containing one or more of the above scene semantic information.

[0066] The semantic analysis model infers user intent based on identified scene semantic information and outputs a correction coefficient for each candidate application based on the correlation between the inferred user intent and the type of each candidate application. Applications that highly match the inferred intent receive a correction coefficient greater than 1 to positively improve their score, while applications that do not match the inferred intent receive a correction coefficient less than 1 to negatively decrease their score. User intent is an inference based on scene semantic information about the user's potential current activity or usage tendency, which may include quiet movie watching, gaming, learning, and exercise. The correction coefficient is a weight adjustment factor output by the semantic analysis model for each candidate application, used to refine the behavior score in the first stage based on the current context, making the final score more closely reflect the user's actual usage needs.

[0067] In one embodiment, the current user is a registered parent user, and the candidate application list includes: live sports, movie streaming, online education, short videos, and children's animation. The system inputs the current living room scene captured by the camera and the aforementioned candidate application list into a semantic analysis model. The semantic analysis model analyzes the scene and identifies the following scene semantic information: the person's emotion is relaxed (relaxed facial muscles, slightly upturned corners of the mouth), the object held is a remote control (the user is holding a TV remote control), and the ambient light is dim (only ambient lighting is on in the living room, and the overall lighting is dim). Based on the above scene semantic information, it is inferred that the user's current intention is to quietly watch movie entertainment content: dim lighting is suitable for immersive viewing, the relaxed emotion indicates that the user prefers leisure and entertainment rather than high-intensity activities, and holding a remote control indicates that the user is in a standby state to switch channels at any time. Based on this, the semantic analysis model outputs correction coefficients for each candidate application: Movie playback 1.3 (dim environment and relaxed mood highly match the viewing scenario), sports live broadcast 0.8 (although it is entertainment, the dim and quiet environment does not fully match the atmosphere of intense competition), short videos 0.6 (the user's posture while holding the remote control indicates prolonged immersion rather than short-term scrolling), children's animation 0.5 (no children were detected in the current scene), and online education 0.3 (relaxed state does not match the learning scenario). These correction coefficients will be multiplied by the behavioral score from the first stage in subsequent steps to achieve refined scoring correction based on scene semantics.

[0068] The technical solution of this application uses a semantic analysis model to perform deep scene semantic recognition on environmental perception data, which can understand fine-grained information such as the user's current emotional state, the items held in hand, and the ambient light, thereby accurately inferring the user's real-time intention; and outputs a correction coefficient based on the correlation between the intention and the candidate application, so that the preloading decision has scene perception capability, which significantly improves the accuracy of preloading prediction in irregular scenarios.

[0069] Based on the above technical solution, as an embodiment, the preloading of the target application may include: calling the system process derivation mechanism to create a base process of the target application to realize the preloading of the target application; wherein, in the preloading stage, only the base process is created and user interface resources are not loaded, so that the application can be started based on the already created base process when the application startup instruction is received.

[0070] The system process forking mechanism is a low-level mechanism provided by the operating system that quickly creates new application processes by copying (forking) an existing base process. A typical implementation is the Zygote process forking mechanism in the Android system. The Zygote process is a base process pre-initialized by the Android system at startup. It pre-loads the common runtime environment and shared libraries required for application operation. When an application needs to be launched, the system quickly creates the base process for that application by forking the Zygote process. The new process inherits the common resources already loaded in the Zygote process, without needing to initialize the runtime environment from scratch. The base process is the core process framework upon which the target application depends for operation. It contains the application's runtime environment, basic libraries, and service connections, but does not include user interface-related resource files.

[0071] The preloading phase refers to the time window from when the system decides to preload the target application to when the user actually initiates the launch command. User interface resources refer to the front-end display-related data used by the application to render the interface during runtime, such as layout files, image assets, animation resources, and UI component instances. In practice, during the preloading phase, the system only creates the basic process of the target application through a fork mechanism, allowing this process to run in the background with a small amount of memory, but actively skips the loading of user interface resources. This ensures that the memory usage during the preloading phase is significantly lower than during a full startup, effectively controlling the consumption of system memory resources by preloading.

[0072] When a user clicks the target application icon or otherwise initiates an application launch command, the system detects that the target application's basic process has already been created during the preloading phase. It can then directly load user interface resources onto the existing basic process and complete the application's full launch. Because the runtime environment and shared libraries are already ready during the preloading phase, the time spent initializing processes from scratch during a cold start is eliminated, significantly improving application startup speed.

[0073] In one embodiment, the video playback application is the target application that needs to be preloaded. The system calls the Android operating system's Zygote process forking mechanism, sending a fork request to the Zygote process. The Zygote process creates a new child process through the fork system call. This child process inherits the preloaded Android runtime (ART), core libraries, and system services from Zygote, becoming the base process of the "video playback" application. At this point, the base process is already running in the background, consuming approximately 30% to 40% of the memory used during a full startup, because user interface resources such as layout files, high-definition images, and video decoders have not yet been loaded.

[0074] A few seconds later, the user clicks the video playback application icon on the remote control, and the system receives the application launch command. The system detects that the basic process for video playback has already been created in the pre-loading phase, and then loads user interface resources on that process: parsing the layout XML file, rendering the main interface, and initializing the video player component. Since the runtime environment is already ready, the interface loading process does not need to wait for underlying initialization, and the application takes only about 0.8 seconds from click to the main interface being fully displayed. In contrast, if a cold start method is used without pre-loading, the system needs to initialize the process from scratch, load the runtime environment, and then load the user interface resources, which takes about 2.5 seconds. The pre-loading solution reduces the startup time by approximately 68%.

[0075] The technical solution adopted in this application's embodiments creates only a basic process without loading user interface resources during the preloading phase through a system process spawning mechanism, achieving application pre-start preparation with lower memory overhead. When the user initiates a startup command, the interface is quickly loaded based on the existing basic process, significantly shortening the startup time compared to a cold start, while avoiding the waste of memory resources caused by full preloading.

[0076] Figure 2 This is a flowchart illustrating an application preloading method provided in an embodiment of this application; see reference. Figure 2 The application preloading method in this application adopts a two-layer architecture, including a base layer and an enhancement layer. After coarse screening in the base layer and fine screening in the enhancement layer, the preloading of the target application is finally completed.

[0077] The system collects current environmental data through cameras on smart terminal devices, and the collected data is processed in the base layer. In the base layer, face detection is performed on the collected images to determine if a face is detected. If a face is detected, the system extracts facial features and matches them with facial feature templates of registered users to determine the current user's identity. After confirming the user's identity, the system further determines whether the user's historical behavior data can be found. If historical behavior data is found, it means the current user is a registered user with usage records. The system applies a time decay model formula to assign a base score to each historical usage behavior of each candidate application based on the user's historical behavior data, and then performs a weighted summation based on time intervals to calculate the base behavior score for each candidate application. If no historical behavior data is found, it means the current user is a newly registered user or a first-time user. The system can identify the current user's age group characteristics and match the behavior scores of each candidate application corresponding to that age group from a general preference knowledge base to achieve initial score coverage in cold start scenarios.

[0078] The system detects whether multiple users are present simultaneously in the current frame (i.e., user groups). If a user group exists, it determines a scenario coefficient based on the group's historical shared usage records and adjusts the weight of candidate applications in the corresponding category accordingly. The system weights the basic behavior score with the scenario coefficient to obtain the behavior score of each candidate application, sorts them from highest to lowest score, and outputs a Top-N candidate application list, completing the coarse screening process at the basic layer.

[0079] In the enhancement layer, current environmental perception data (such as images of the current environment captured by a camera) is collected. This environmental perception data, along with information from the Top-N candidate applications, is input into a semantic analysis model deployed on the device. The semantic analysis model performs scene semantic recognition on the environmental perception data, identifying fine-grained semantic information such as human emotions, held objects, and ambient lighting. Based on this, it infers the user's current real-time intent and outputs corresponding correction coefficients for each candidate application. Subsequently, the system performs a final weighted calculation (such as multiplication) on the correction coefficients and the behavior scores obtained by each candidate application in the base layer to obtain the final score for each candidate application. The candidate application with the highest score is determined as the target application, and a preloading operation is performed on the determined target application. Specifically, the system calls the system process forking mechanism (such as the fork mechanism of the Android Zygote process) to create the base process of the target application. During the preloading stage, only the process framework is created without loading user interface resources. This allows the application to start quickly and completely based on the already created base process when the user subsequently initiates the application launch command, significantly shortening the startup waiting time.

[0080] Thus, through a two-layer architecture of a base layer and an enhancement layer, the base layer quickly completes the coarse screening of candidate applications with lightweight computing, reducing the data scale of subsequent processing; the enhancement layer only performs semantic analysis and fine ranking on the screened candidate applications, avoiding the computational overhead of full AI inference. This two-layer progressive processing enables deep intelligent decision-making while ensuring the real-time performance of preloading, making it particularly suitable for intelligent terminal devices with limited computing power.

[0081] To facilitate better implementation of the application preloading method of this application, this application also provides an application preloading apparatus based on the above-described application preloading method. The meanings of the terms used are the same as in the application preloading method described above, and specific implementation details can be found in the descriptions of the method embodiments.

[0082] Please see Figure 3 , Figure 3 This is a schematic diagram of the application preloading device provided in an embodiment of this application, wherein the application preloading device includes: The acquisition module 301 is used to acquire the historical behavior data of the current user and determine the behavior score of each candidate application based on the historical behavior data; The determination module 302 is used to acquire current environmental perception data and determine the correction coefficient of each of the candidate applications based on the current environmental perception data and the information of multiple candidate applications. The correction module 303 is used to correct the behavior score of each candidate application based on the correction coefficient to obtain the final score of each candidate application. The preloading module 304 is used to determine the target application based on the final score and preload the target application.

[0083] In one embodiment, the acquisition module 301 is specifically used to perform: Based on the time decay model, a basic score is assigned to each historical usage behavior of each candidate application. The weight of each basic score is determined based on the interval between each historical usage behavior of each candidate application and the current time; wherein the weight of the historical usage behavior that is closer to the current time is greater. The basic scores of each candidate application's historical usage behavior are weighted and summed based on the weights to obtain the behavioral score of each candidate application.

[0084] In one embodiment, when the current user is a user group, the device further includes: The record acquisition module is used to acquire the historical usage records of each of the candidate applications based on user combination information; The scenario coefficient determination module is used to determine the scenario coefficient based on the historical usage records. The weighted summation of the baseline scores for each historical usage behavior of each candidate application based on the weights, to obtain the behavioral score of each candidate application, includes: Based on the weights, the basic scores of each candidate application's historical usage behavior are weighted and summed to obtain the basic behavior score of each candidate application. The basic behavioral scores of each candidate application are weighted based on the scenario coefficients to obtain the behavioral scores of each candidate application.

[0085] In one embodiment, the correction module 303 is specifically configured to perform: The current environmental perception data and information from multiple candidate applications are input into the semantic analysis model; The semantic analysis model identifies scene semantic information from the current environmental perception data; the scene semantic information includes one or more of the following: human emotions, handheld objects, and ambient lighting. Based on the semantic information of the scenario, the user intent is inferred, and the correction coefficient of each of the candidate applications is determined according to the user intent.

[0086] In one embodiment, the acquisition module 301 is specifically used to perform: Collect the facial data of the current user, and extract the facial features of the current user based on the facial data; The facial features are matched with the facial features of registered users to determine the identity of the current user; Obtain the historical behavior data corresponding to the identity identifier.

[0087] In one embodiment, the device further includes: An age recognition module is used to identify the age characteristics of the current user when the facial feature vector does not match the facial features of a registered user. The rating search module is used to find the behavioral ratings of each candidate application corresponding to the age characteristics.

[0088] In one embodiment, the preloading module 304 is specifically used to perform: The system process spawning mechanism is invoked to create the base process of the target application in order to preload the target application; In the preloading phase, only the basic process is created without loading user interface resources, so that the application can be launched based on the already created basic process when the application launch instruction is received.

[0089] The technical solution adopted in this application determines the behavioral scores of each candidate application by acquiring the historical behavioral data of the current user. This provides an accurate user-side data foundation for preloading decisions based on the actual usage of the current user rather than all users, ensuring that the target application for preloading matches the current user's preferences. Acquiring current environment awareness data and combining it with candidate application information to determine correction coefficients enables the system to perceive the current usage scenario and capture real-time intent information that cannot be obtained from historical data alone. The behavioral scores are corrected based on the correction coefficients to obtain the final score, effectively integrating long-term user preferences with real-time environment awareness. This ensures that the final score reflects both user usage patterns and adapts to the current usage scenario, significantly improving the accuracy of preloading prediction. Determining the target application based on the final score and preloading it ensures that the application most suitable for the user's current needs is preloaded, improving application startup speed while avoiding memory resource waste and system performance degradation caused by blind preloading.

[0090] Specific limitations regarding the application preloading device can be found in the limitations of the application preloading method above, and will not be repeated here. Each module in the aforementioned application preloading device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in the computer device, or stored in software in the memory of the computer device, so that the processor can call and execute the operations corresponding to each module.

[0091] In addition, this application also provides an electronic device, such as Figure 4 As shown, it illustrates the structural diagram of the electronic device involved in this application, specifically: The electronic device may include components such as a processor 401 with one or more processing cores and a memory 402 with one or more computer-readable storage media. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein: The processor 401 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0092] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0093] In one embodiment, the electronic device further includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power equipment debugging circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0094] In one embodiment, the electronic device may further include an input unit 404, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0095] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402, thereby implementing the steps in any of the application preloading methods provided in the embodiments of this application.

[0096] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0097] In one embodiment, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the methods described in any embodiment of this application.

[0098] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of this application.

[0099] In some embodiments, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the methods described in any embodiment of this application.

[0100] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0101] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0102] Therefore, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor to perform the steps in any of the application preloading methods provided in this application.

[0103] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0104] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0105] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the application preloading methods provided in this application, the beneficial effects that any of the application preloading methods provided in this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0106] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0107] The above provides a detailed description of an application preloading method, apparatus, electronic device, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An application preloading method, characterized in that, include: Obtain the current user's historical behavior data, and determine the behavior score of each candidate application based on the historical behavior data; Acquire current environmental perception data, and based on the current environmental perception data and information of multiple candidate applications, determine the correction coefficient for each candidate application; The behavioral scores of each candidate application are corrected based on the correction coefficient to obtain the final scores of each candidate application. The target application is determined based on the final score, and the target application is preloaded.

2. The method according to claim 1, characterized in that, The process of determining the behavioral score of each candidate application based on the historical behavioral data includes: Based on the time decay model, a basic score is assigned to each historical usage behavior of each candidate application. The weight of each basic score is determined based on the interval between each historical usage behavior of each candidate application and the current time; wherein the weight of the historical usage behavior that is closer to the current time is greater. The basic scores of each candidate application's historical usage behavior are weighted and summed based on the weights to obtain the behavioral score of each candidate application.

3. The method according to claim 2, characterized in that, When the current user is a user group, the method further includes: Obtain the historical usage records of each of the candidate applications based on user combination information; Determine the scenario coefficient based on the historical usage records; The weighted summation of the baseline scores for each historical usage behavior of each candidate application based on the weights, to obtain the behavioral score of each candidate application, includes: Based on the weights, the basic scores of each candidate application's historical usage behavior are weighted and summed to obtain the basic behavior score of each candidate application. The basic behavioral scores of each candidate application are weighted based on the scenario coefficients to obtain the behavioral scores of each candidate application.

4. The method according to claim 1, characterized in that, The step of determining the correction coefficient for each of the candidate applications based on the current environmental perception data and information from multiple candidate applications includes: The current environmental perception data and information from multiple candidate applications are input into the semantic analysis model; The semantic analysis model identifies scene semantic information from the current environmental perception data; the scene semantic information includes one or more of the following: human emotions, handheld objects, and ambient lighting. Based on the semantic information of the scenario, the user intent is inferred, and the correction coefficient of each of the candidate applications is determined according to the user intent.

5. The method according to any one of claims 1 to 4, characterized in that, The acquisition of the current user's historical behavior data includes: Collect the facial data of the current user, and extract the facial features of the current user based on the facial data; The facial features are matched with the facial features of registered users to determine the identity of the current user; Obtain the historical behavior data corresponding to the identity identifier.

6. The method according to claim 5, characterized in that, The method further includes: When the facial feature vector does not match the facial features of a registered user, the age feature of the current user is identified; Find the behavioral scores of each candidate application corresponding to the age feature.

7. The method according to any one of claims 1 to 4, characterized in that, The preloading of the target application includes: The system process spawning mechanism is invoked to create the base process of the target application in order to preload the target application; In the preloading phase, only the basic process is created without loading user interface resources, so that the application can be launched based on the already created basic process when the application launch instruction is received.

8. An application preloading device, characterized in that, include: The acquisition module is used to acquire the current user's historical behavior data and determine the behavior score of each candidate application based on the historical behavior data. The determination module is used to acquire current environmental perception data and, based on the current environmental perception data and information of multiple candidate applications, determine the correction coefficient of each candidate application. The correction module is used to correct the behavior scores of each candidate application based on the correction coefficient to obtain the final scores of each candidate application. A preloading module is used to determine the target application based on the final score and preload the target application.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the application preloading method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the application preloading method as described in any one of claims 1 to 7.

Images