An application preheating terminal system based on artificial intelligence

Abstract

The application discloses an application preheating terminal system based on artificial intelligence and belongs to the technical field of application preheating. In the system, a full-amount data analysis module collects application use data of all authorized users as full-amount data, carries out joint analysis on the full-amount data based on CNN and LSTM, obtains general use habits and general use trends for all authorized users, and generates general preheating strategies; a personalized data training module trains a personalized preheating model according to the use data of the user on the application, the personalized preheating model optimizes the use data of the user and predicts the use habits of the user on the application; and a full-amount data and personalized data fusion module fuses the general preheating strategies and the predicted use habits of the user on the application to preheat the application. The application combines large-scale global data and personalized end-side models, so that the scheduling of preheating resources is more accurate.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of application preheating, and particularly relates to an application preheating terminal system based on artificial intelligence. BACKGROUND

[0002] In modern mobile terminals, in order to accelerate the application startup speed, a preheating technology is usually used, that is, part of resources of the application, such as dependent dynamic libraries and other commonly used components, are loaded in advance. The traditional preheating method is usually based on general rules and cannot be optimized according to the use habits of each user. However, the traditional preheating method based on general rules cannot be optimized according to the use habits of each user. This may cause waste of part of preheating resources, or the application is not fully preheated when starting, which affects the system performance and user experience. SUMMARY

[0003] In view of the above technical problems, the application provides an application preheating scheme based on artificial intelligence.

[0004] The first aspect of the application provides an application preheating terminal system based on artificial intelligence, which comprises: a full-data analysis module located at a cloud side, and a personalized data training module, a full-data and personalized data fusion module located at a terminal side; wherein:

[0005] At the cloud side, the full-data analysis module is configured to: collect application use data of all authorized users as full data, perform joint analysis based on CNN and LSTM on the full data, obtain general use habits and general use trends facing all authorized users, and generate a general preheating strategy;

[0006] At the terminal side, the personalized data training module is configured to: train a personalized preheating model according to the use data of the user on the application, the personalized preheating model performs data optimization on the use data of the user and predicts the use habits of the user on the application;

[0007] At the terminal side, in response to the terminal starting the preheating process of the application, the full-data and personalized data fusion module is configured to: receive the general preheating strategy, call the trained personalized preheating model to predict the use habits of the user on the application, and fuse the general preheating strategy and the predicted use habits of the user on the application to preheat the application.

[0008] According to the system of the first aspect of the application, at the cloud side, the full-data analysis module is configured to: collect application use data of all authorized users as full data; specifically comprising:

[0009] The system records the startup time when the user launches the application and the shutdown time when the application enters the background. It collects user application usage data within a preset time interval and performs anonymization processing on the data. During idle periods, the anonymized application usage data is uploaded to the cloud as the full dataset.

[0010] According to the system of the first aspect of the present invention, on the cloud side, the full data analysis module is configured to: perform joint analysis of the full data based on CNN and LSTM to obtain general usage habits and trends for all authorized users, and generate a general preheating strategy; specifically including:

[0011] By using CNN and LSTM to perform a comprehensive analysis of the full dataset, the application's launch patterns and resource requirements are extracted. Furthermore, an application launch heatmap is generated, from which common usage habits and trends for all authorized users are obtained. A common preheating strategy is then generated and distributed to the terminal at a predetermined time.

[0012] According to the system of the first aspect of the present invention, on the cloud side:

[0013] A hybrid model is constructed using CNN and LSTM; the convolutional neural network (CNN) extracts spatial features from the sequence data and inputs the spatial features into the long short-term memory network (LSTM) to capture long-term dependencies in the temporal dimension.

[0014] When training the hybrid model, the training dataset is represented as , This represents the i-th input sequence data. This represents the preheating hit rate achieved by the preheating strategy generated based on the i-th set of sequence data, where N represents the number of sequence data.

[0015] According to the system of the first aspect of the present invention, on the terminal side, the personalized data training module is configured to: train a personalized preheating model based on user usage data of the application; wherein:

[0016] The training process employs incremental learning, and the model parameters for the personalized preheating model are as follows: The collected user usage data for the application is , This represents the user's i-th set of usage data for the application. M represents the preheating hit rate predicted based on the user's usage habits of the application based on the i-th set of user usage data for the application, and M represents the amount of user usage data for the application.

[0017] The model parameters of the personalized warm-up model are updated using stochastic gradient descent.

[0018] ,

[0019] wherein, denotes the updated model parameters, and η denotes the learning rate, denotes the loss function.

[0020] According to the system of the first aspect of the application, on the terminal side, the personalized preheating model prediction of the user's use habit of the application includes: multi-application switching frequency, commonly used time period, and geographical location.

[0021] According to the system of the first aspect of the application, on the terminal side, in response to the terminal starting the preheating process of the application, the full-data and personalized data fusion module is configured to:

[0022] receive a general preheating strategy, and obtain features extracted from the cloud ;

[0023] invoke the trained personalized preheating model to predict the user's use habit of the application, and obtain features extracted from the terminal ;

[0024] fuse the general preheating strategy and the predicted user's use habit of the application, and splice and into a comprehensive feature vector ; wherein:

[0025] ,

[0026] Further introduce a weight adjustment mechanism:

[0027] wherein, denotes the weight adjustment coefficient.

[0028] The second aspect of the application provides an application preheating terminal method based on artificial intelligence, which comprises:

[0029] On the cloud side:

[0030] Collect application use data of all authorized users as full-data, perform joint analysis based on CNN and LSTM on the full-data, obtain general use habits and general use trends for all authorized users, and generate a general preheating strategy;

[0031] On the terminal side:

[0032] Train a personalized preheating model according to the user's use data of the application, wherein the personalized preheating model optimizes the user's use data and predicts the user's use habit of the application;

[0033] A general preheating strategy is received, a trained personalized preheating model is called to predict a user's usage habit for an application, and the general preheating strategy and the predicted usage habit of the user for the application are fused to preheat the application.

[0034] The third aspect of the present application discloses an electronic device. The electronic device comprises a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the application preheating terminal system based on artificial intelligence of the second aspect of the present application is realized.

[0035] The fourth aspect of the present application discloses a computer readable storage medium. The computer readable storage medium stores a computer program. When the computer program is executed by a processor, the application preheating terminal system based on artificial intelligence of the second aspect of the present application is realized.

[0036] In summary, the scheme provided by the present application adopts a deep learning algorithm, combines application usage data of a large number of authorized users and personalized usage behaviors of a single user, and intelligently and customizes preheating of an application on a mobile terminal. This way not only improves the preheating hit rate, but also effectively reduces energy consumption and improves user experience. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can obtain other drawings without creative labor on the basis of these drawings.

[0038] Figure 1 The application preheating flowchart according to the embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0040] The first aspect of the present application proposes an application preheating terminal system based on artificial intelligence. The system comprises a full data analysis module located on the cloud side, and a personalized data training module and a full data and personalized data fusion module located on the terminal side. As shown in the figure: Figure 1

[0041] ​On the cloud side, the full-data analysis module is configured to: collect application usage data of all authorized users as full data, perform joint analysis on the full data based on CNN and LSTM to obtain general usage habits and general usage trends for all authorized users, and generate a general preheating strategy;

[0042] On the terminal side, the personalized data training module is configured to: train a personalized preheating model according to the usage data of the user on the application, wherein the personalized preheating model optimizes the usage data of the user and predicts the usage habits of the user on the application;

[0043] On the terminal side, in response to the terminal starting the preheating process of the application, the full-data and personalized-data fusion module is configured to: receive the general preheating strategy, call the trained personalized preheating model to predict the usage habits of the user on the application, and fuse the general preheating strategy and the predicted usage habits of the user on the application to preheat the application.

[0044] In some embodiments, on the cloud side, the full-data analysis module is configured to: collect application usage data of all authorized users as full data; specifically including:

[0045] When the user starts the application, record the start time, when the application enters the background, record the stop time, collect the application usage data of the user within a preset time interval and perform desensitization processing, and upload the desensitized application usage data to the cloud as full data in idle time.

[0046] In some embodiments, on the cloud side, the full-data analysis module is configured to: perform joint analysis on the full data based on CNN and LSTM to obtain general usage habits and general usage trends for all authorized users, and generate a general preheating strategy; specifically including:

[0047] Using CNN and LSTM to comprehensively analyze the full data, extracting the start mode and resource demand of the application, further generating an application start heat map, obtaining general usage habits and general usage trends for all authorized users from the heat map, and generating a general preheating strategy, and issuing the general preheating strategy to the terminal at a predetermined time.

[0048] In some embodiments, on the cloud side:

[0049] A hybrid model is constructed using CNN and LSTM; the convolutional neural network CNN extracts spatial features from sequence data, and the spatial features are input to the long short-term memory network LSTM to capture long-term dependencies in the time dimension;

[0050] When training the hybrid model, the training data set is represented as , represents the i-th set of input sequence data, represents the preheat hit rate reached by the preheat strategy generated based on the i-th set of sequence data, N represents the number of sequence data.

[0051] Specifically, for full data analysis, the system avoids the problem of cold start without data by collecting application usage data of all authorized users. The cloud dataset comprehensively analyzes a large number of applications, extracts the startup mode and resource demand of various applications, and thus ensures the basic accuracy of preheating at the big data level.

[0052] Specifically, high-performance collection of authorized user application usage data adopts the strategy of busy time recording and idle time reporting to achieve the goal of high-performance collection. Specifically, when the authorized user starts the application, the start time is recorded, and when the application enters the background, the stop time is recorded. Through this method, the application usage record of the current authorized user within 24 hours is recorded, including start time, single use duration, total use duration, start times, and other multi-dimensional data. In the idle time (night period, mobile phone connected to charging and wifi state), the data is desensitized, the user identity sensitive information is ignored, and then uploaded to the cloud.

[0053] Specifically, the cloud analyzes the full application usage habits, and analyzes the behavior habits of all authorized users through convolutional neural network CNN and long short-term memory network LSTM. We use a scoring method, that is, define key indicators (such as application start times, use duration) and assign weights, and quantify user behavior into scores, and then generate an application startup heat map to analyze the overall usage habits and trends of the application. The trained cloud strategy will be issued to the end side at a specific time period (such as system update), and will be updated regularly. The end side will preheat the application according to the full data when cold starting.

[0054] Specifically, the entire system architecture draws on the idea of federated learning. Each participant independently trains the model using local data and uploads the desensitized data to the central server. The server aggregates information from all parties to optimize the global model and distributes the updated model to each participant. This continuous iteration and optimization forms an efficient and secure closed-loop learning system.

[0055] The cloud collects application usage data from a large number of authorized terminals, covering application start time, stop time, use duration, and other multi-dimensional information.

[0056] Among them, the preheat hit rate refers to the ratio of the number of times a user uses a preheat application to the total number of preheat applications within a certain period of time, and its calculation formula is:

[0057]

[0058] For the training data set , represents the inputted i-th group of sequence data, represents the preheat hit rate reached by the preheat strategy generated based on the i-th group of sequence data, and N represents the number of sequence data.

[0059] After the data collection is completed, a data cleaning algorithm is used in combination with context information such as time zone and user behavior patterns to accurately determine the validity of the data, thereby reducing the situation of mistakenly deleting valid data. Subsequently, the data is normalized based on distribution to ensure comparability and consistency of the data.

[0060] In actual training, it is found that the preheat hit rate not only has local features and spatial correlation with the frequency and habits of users using the application, but also presents a time trend and dependency relationship. For example, after using application A, the user will often use application B immediately. In order to comprehensively consider the discrete feature points and time dependency relationship, the cloud uses a hybrid model constructed by CNN and LSTM algorithms. Among them, the convolutional neural network CNN can efficiently process the sequence data and extract the spatial features therein. After extracting the spatial features through convolution operation, the result is input into the long short-term memory network LSTM to capture the long-term dependency relationship in the time dimension, and then the application usage patterns of the user are more comprehensively and accurately mined.

[0061] In some embodiments, on the terminal side, the personalized data training module is configured to train a personalized preheat model according to the use data of the user on the application; wherein:

[0062] The training process adopts an incremental learning manner, and the model parameters of the personalized preheat model are , and the collected use data of the user on the application is , represents the i-th group of use data of the user on the application, represents the preheat hit rate reached by the use habits of the user on the application predicted based on the i-th group of use data of the user on the application, and M represents the number of use data of the user on the application;

[0063] The model parameters of the personalized preheat model are updated by the stochastic gradient descent method:

[0064]

[0065] wherein, represents the updated model parameters, η represents the learning rate, represents the loss function.

[0066] In some embodiments, on the terminal side, the use habits of the user on the application predicted by the personalized preheat model include: multi-application switching frequency, common time period, and geographic location.

[0067] Specifically, for multi-dimensional personalized model training: the end-side system trains a personalized pre-warming model according to the user's device usage habits, such as multi-application switching frequency, commonly used time period, geographic location, and historical pre-warming hit rate, and other multi-dimensional parameters. The model can be optimized for different user usage habits to ensure that pre-warming resources can effectively cover actual needs.

[0068] Specifically, the end-side predicts the target user's application usage habits, uses the system's built-in lightweight training model, combines user-side user personal data, and combines user usage habits to perform personalized portrait analysis on user behavior and predict and analyze the target user's usage habits. This part of personal data is only used on the end-side and will not be uploaded to the cloud, ensuring personal information security.

[0069] Specificly, the end-side training first performs data collection and preprocessing. The end-side mainly collects local user's personalized application usage data. When preprocessing the data, a more refined data simplification strategy is adopted, such as sampling combined with data importance evaluation. According to the importance of application usage data to the pre-warming strategy, the data is sampled by level, the key data is retained, and the storage and calculation pressure of unimportant data is reduced.

[0070] Specifically, according to the usage habits and data of local users, the model is trained. Incremental learning is adopted, and the existing model parameters of the local user are collected , represents the user's i-th set of usage data for the application, represents the pre-warming hit rate reached by the user's usage habits for the application based on the user's i-th set of usage data for the application, and M represents the number of usage data of the user for the application.

[0071] The model parameters are updated by stochastic gradient descent:

[0072]

[0073] wherein, represents the updated model parameters, η represents the learning rate, represents the loss function.

[0074] In some embodiments, on the terminal side, in response to the terminal starting the pre-warming process for the application, the full data and personalized data fusion module is configured to:

[0075] receive a general pre-warming strategy, and obtain features extracted from the cloud ;

[0076] invoke the trained personalized pre-warming model to predict the user's usage habits for the application, and obtain features extracted from the terminal ;

[0077] Fusion general preheating strategy and predicted user application usage habits, will And Spliced into a comprehensive feature vector ;

[0078] Among them:

[0079] Further introduce weight adjustment mechanism:

[0080] Among them, Indicates the weight adjustment coefficient.

[0081] Specifically, for the combination of full data and personalized data: the system combines full user data and single user personalized use behavior to form a more accurate preheating strategy. This combination can significantly improve the preheating hit rate, reduce unnecessary resource loading, and further improve the response speed of the device and user experience.

[0082] Specifically, the cloud data and personalized data are combined to predict application usage habits, and the strategy is issued in the form of input into the end-side prediction training set. The end-side will combine cloud data and end-side prediction analysis to more comprehensively and customize the prediction and analysis of user application usage habits. This method can fully utilize the global data of the cloud and the personalized data of the end-side to improve the accuracy of prediction and the efficiency of application preheating.

[0083] Specifically, for cloud and end-side feature fusion:

[0084] Among them, for feature-level fusion:

[0085] After the cloud trains the full feature model, it first compresses the model through quantization, pruning, knowledge distillation, etc., and then issues it to the device end. At the same time, the cloud uses feature selection algorithms (such as information gain, chi-square test) to screen out key features for preheating strategy and issue them.

[0086] The feature vector extracted by the cloud , the local feature vector extracted at the end side , the And Spliced into a comprehensive feature vector : .

[0087] The spliced feature vector is used as input for the subsequent preheating strategy prediction model. By introducing a weight adjustment mechanism, the fusion weight is adjusted according to the current data features. The weighted fused feature vector can be represented as:

[0088] wherein, ∈[0,1] is a weight adjustment coefficient.

[0089] Specifically, taking the daily application usage time and the launch frequency of the user as an example, the feature vector extracted by the cloud =[120,8] (representing the daily usage time of an application of 120 seconds, and the launch of 8 times, respectively). The local feature vector extracted by the device =[5,1] (such as the installation location code of the application on the device is 5, and the user-defined label is 1). The comprehensive feature vector is obtained . Finally, the spliced feature vector is input into the preheating strategy prediction model to obtain the decision of whether to preheat the application.

[0090] In some embodiments, regarding decision-level fusion: in the manner of decision-level fusion, two models make predictions respectively, and then the prediction results are integrated through a voting mechanism. If the voting result exceeds a predetermined threshold, the operating system will preheat the specified application. Specifically, the end-side device uses the global model features obtained from the cloud and the personalized model features trained locally to make independent predictions. For example, in a classification problem, if the global model on the end-side predicts the category as , and the local model on the end-side predicts the category as , the final prediction category is determined through a voting mechanism. For example: the model predicts that the preheating result of the compressed cloud model is “preheat”. The end-side model that fuses the key gain features predicts “preheat”. In a simple majority voting, both models predict “preheat”, and the voting result exceeds the predetermined threshold, so the operating system will preheat the application.

[0091] The scheme proposed by the present application has the following technical advantages: (1) accuracy: combining large-scale global data and personalized end-side models makes the scheduling of preheating resources more accurate, effectively reducing the problem of not being preheated at startup. (2) efficiency: through intelligent preheating strategies, unnecessary resource preloading is reduced, system energy consumption is significantly reduced, and application startup speed is improved. (3) adaptability: based on the continuous training of deep learning models, the system can continuously adapt to user usage habits and environmental changes, ensuring the long-term effectiveness of the preheating strategy.

[0092] The present application not only improves the startup speed of mobile terminal applications, but also optimizes the utilization rate of system resources through intelligent resource scheduling, reduces power consumption, and greatly improves the overall performance of user experience and device performance.

[0093] The second aspect of the present application proposes an application preheating terminal method based on artificial intelligence, which comprises:

[0094] On the cloud side:

[0095] Collect the application usage data of all authorized users as full data, perform joint analysis on the full data based on CNN and LSTM to obtain general usage habits and general usage trends for all authorized users, and generate a general preheating strategy;

[0096] On the terminal side:

[0097] Train a personalized preheating model according to the usage data of the user on the application, the personalized preheating model optimizes the usage data of the user and predicts the usage habits of the user on the application;

[0098] Receive the general preheating strategy, call the trained personalized preheating model to predict the usage habits of the user on the application, and fuse the general preheating strategy and the predicted usage habits of the user on the application to preheat the application.

[0099] The third aspect of the application discloses an electronic device. The electronic device comprises a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the second aspect of the application is realized.

[0100] The fourth aspect of the application discloses a computer readable storage medium. The computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the second aspect of the application is realized.

[0101] In summary, the scheme provided by the application adopts a deep learning algorithm, combines the application usage data of a large number of authorized users and the personalized usage behavior of a single user, and intelligently and customizes the preheating of the application on the mobile terminal. This way not only improves the preheating hit rate, but also effectively reduces energy consumption and improves user experience.

[0102] Please note that the technical features of the above embodiments can be combined in any way, and in order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the description. The above embodiments only express several embodiments of the application, and the description is more specific and detailed, but it should not be interpreted as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the application, a number of modifications and improvements can be made, which are within the scope of the application. Therefore, the scope of the patent of the application should be subject to the appended claims.

Claims

1. An application preheating terminal system based on artificial intelligence, characterized in that, The system includes: a full-data analysis module located on the cloud side, and a personalized data training module and a full-data and personalized data fusion module located on the terminal side; wherein: On the cloud side, the full data analysis module is configured to: collect application usage data from all authorized users as full data, perform joint analysis of the full data based on CNN and LSTM to obtain general usage habits and trends for all authorized users, and generate a general warm-up strategy; On the terminal side, the personalized data training module is configured to: train a personalized preheating model based on the user's application usage data, wherein the personalized preheating model optimizes the user's usage data and predicts the user's application usage habits; On the terminal side, in response to the application warm-up process when the terminal starts, the full data and personalized data fusion module is configured to: receive the general warm-up strategy, call the trained personalized warm-up model to predict the user's usage habits of the application, and fuse the general warm-up strategy and the predicted user's usage habits of the application to warm up the application. On the terminal side, in response to the application warm-up process upon terminal startup, the full data and personalized data fusion module is configured as follows: Receive a general preheating strategy and obtain features extracted from the cloud. ; The trained personalized warm-up model is invoked to predict user usage habits of the application and to obtain features extracted from the terminal. Integrating general preheating strategies and predicted user habits for the application, and Concatenate into a comprehensive feature vector ;in: ； Further introduce a weight adjustment mechanism: ; in, Indicates the weight adjustment factor; The concatenated feature vector is used as input for the subsequent warm-up strategy prediction model.

2. The application preheating terminal system based on artificial intelligence according to claim 1, characterized in that, On the cloud side, the full data analysis module is configured to collect application usage data from all authorized users as the full dataset; specifically, this includes: The system records the startup time when the user launches the application and the shutdown time when the application enters the background. It collects user application usage data within a preset time interval and performs anonymization processing on the data. During idle periods, the anonymized application usage data is uploaded to the cloud as the full dataset.

3. The application preheating terminal system based on artificial intelligence according to claim 2, characterized in that, On the cloud side, the full data analysis module is configured to perform joint analysis of the full data based on CNN and LSTM to obtain general usage habits and trends for all authorized users, and generate a general warm-up strategy; specifically including: By using CNN and LSTM to perform a comprehensive analysis of the full dataset, the application's launch patterns and resource requirements are extracted. Furthermore, an application launch heatmap is generated, from which common usage habits and trends for all authorized users are obtained. A common preheating strategy is then generated and distributed to the terminal at a predetermined time.

4. The application preheating terminal system based on artificial intelligence according to claim 3, characterized in that, On the cloud side: A hybrid model is constructed using CNN and LSTM; the convolutional neural network (CNN) extracts spatial features from the sequence data and inputs the spatial features into the long short-term memory network (LSTM) to capture long-term dependencies in the temporal dimension. When training the hybrid model, the training dataset is represented as , This represents the i-th input sequence data. This represents the preheating hit rate achieved by the preheating strategy generated based on the i-th set of sequence data, where N represents the number of sequence data.

5. The application preheating terminal system based on artificial intelligence according to claim 4, characterized in that, On the terminal side, the personalized data training module is configured to: train a personalized preheating model based on user usage data of the application; wherein: The training process employs incremental learning, and the model parameters for the personalized preheating model are as follows: The collected user usage data for the application is , This represents the user's i-th set of usage data for the application. M represents the preheating hit rate predicted based on the user's usage habits of the application based on the i-th set of user usage data for the application, and M represents the amount of user usage data for the application. The model parameters of the personalized warm-up model are updated using stochastic gradient descent. ； in, This represents the updated model parameters, where η represents the learning rate. This represents the loss function.

6. The application preheating terminal system based on artificial intelligence according to claim 5, characterized in that, On the terminal side, the personalized preheating model predicts users' application usage habits, including: frequency of switching between multiple applications, frequently used time periods, and geographical location.

7. A method for preheating an application terminal based on artificial intelligence, characterized in that, The method includes: On the cloud side: The application usage data of all authorized users is collected as the full dataset. The full dataset is then subjected to joint analysis based on CNN and LSTM to obtain the general usage habits and trends of all authorized users, and to generate a general warm-up strategy. On the terminal side: A personalized preheating model is trained based on user usage data of the application. The personalized preheating model optimizes the user usage data and predicts the user's usage habits of the application. It receives a general preheating strategy, calls a trained personalized preheating model to predict users’ usage habits of the application, and integrates the general preheating strategy and the predicted user usage habits of the application to preheat the application. On the terminal side, in response to the application warm-up process upon terminal startup, the full data and personalized data fusion module is configured as follows: Receive a general preheating strategy and obtain features extracted from the cloud. ; The trained personalized warm-up model is invoked to predict user usage habits of the application and to obtain features extracted from the terminal. ; By integrating general pre-heating strategies and predicted user habits for the application, and Concatenate into a comprehensive feature vector ;in: ; Further introduce a weight adjustment mechanism: ; in, Indicates the weight adjustment factor; The concatenated feature vector is used as input for the subsequent warm-up strategy prediction model.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the application preheating terminal method based on artificial intelligence as described in claim 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the application warm-up terminal method based on artificial intelligence as described in claim 7.

Images