Terminal equipment performance optimization method, device and equipment, and training method, device and equipment of performance optimization model of terminal equipment

Abstract

The invention provides a terminal equipment performance optimization method, device and equipment, and a training method, device and equipment of a performance optimization model of the terminal equipment. The method comprises the following steps: acquiring running data of a plurality of application programs installed in the terminal equipment in real time; inputting the operation data to the performance optimization model, and obtaining a scheduling strategy used for predicting the operation of the application program; generating an application program scheduling instruction corresponding to the scheduling strategy; the performance optimization model is obtained based on training of a plurality of samples, and the samples comprise historical operation data of the application program at a plurality of moments and application program control strategies of continuous time periods associated with the historical operation data. Therefore, the defects of a pre-designed application program scheduling strategy with fixed priority ranking in the prior art are overcome, the accuracy and the flexibility of scheduling strategy prediction are improved, the accuracy and the flexibility of application program scheduling instruction generation are further improved, the performance of terminal equipment is optimized, and the use experience and the satisfaction degree of a user are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device and equipment for optimizing terminal device performance and training a performance optimization model of a terminal device. Background Art

[0002] In the prior art, intelligent systems of terminal devices such as car computers and mobile phones usually use pre-set fixed algorithms to optimize the performance of terminal devices. In other words, development engineers need to manually design the priority sorting rules for each application installed in the terminal device, generate corresponding codes based on the priority sorting rules, and run the code logic through the intelligent system to achieve the balance between applications.

[0003] It can be seen that existing intelligent systems often require manual design and maintenance by development engineers when optimizing terminal performance. This not only increases development costs and workload, but also the pre-designed fixed priority sorting may be difficult to meet users' personalized expectations and habits, thereby affecting user experience and acceptance, and reducing user experience and satisfaction. Summary of the invention

[0004] The present invention provides a method, device and equipment for optimizing terminal device performance and training a performance optimization model of a terminal device, so as to solve the defects in the prior art that when an intelligent system optimizes terminal performance, the development cost and workload are large, and the pre-designed fixed priority sorting may be difficult to meet the personalized expectations and habits of users, thereby affecting the user experience and acceptance, and reducing the user experience and satisfaction.

[0005] The present invention provides a method for optimizing terminal device performance, comprising the following steps: Obtain real-time operation data of multiple applications installed in terminal devices; Inputting the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; generating an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained based on training of multiple samples, wherein the samples include historical operation data of the application at multiple times, and application control strategies for successive time periods associated with the historical operation data.

[0006] According to a terminal device performance optimization method provided by the present invention, the operation data includes application data and associated data, and the inputting of the operation data into a performance optimization model to obtain a scheduling strategy for predicting application program operation includes: The operation data is input into the performance optimization model, the performance optimization model generates the predicted probability of the application, and the scheduling strategy is determined based on the predicted probability of the application.

[0007] According to a terminal device performance optimization method provided by the present invention, determining the scheduling strategy based on the predicted probability of the application program includes: Based on the performance optimization model, selecting a target prediction probability from the prediction probabilities of the application; A target application corresponding to the target prediction probability is determined, and the scheduling strategy is optimized based on a program interface type corresponding to the target application.

[0008] According to a terminal device performance optimization method provided by the present invention, the method of generating a corresponding application scheduling instruction based on the scheduling strategy also includes: In response to a user's feedback operation, the feedback operation is compared with the scheduling strategy to obtain a prediction accuracy rate and a number of incorrect predictions; the feedback operation includes manually launching an application and manually killing an application; The scheduling strategy is optimized based on the prediction accuracy, the number of incorrect predictions and the operating data.

[0009] The present invention also provides a method for training a performance optimization model of a terminal device, comprising: Acquire historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; Obtaining an application control policy label for a successive period corresponding to each historical operation data; Generate training samples based on the historical operation data and the application control policy labels corresponding to the historical operation data; The performance optimization model is trained based on the training samples.

[0010] According to a method for training a performance optimization model of a terminal device provided by the present invention, the step of determining the historical association data includes: Acquire user switching behavior data between the multiple applications; Based on the switching behavior data, a correlation matrix is ​​constructed; the off-diagonal elements in the correlation matrix represent the number of switching between two applications, and the diagonal elements in the correlation matrix are zero; Determining the historical association data based on the values ​​of each off-diagonal element in the association matrix; The historical association data indicates the relative frequency of switching between two application programs associated with the element.

[0011] According to a training method for a performance optimization model of a terminal device provided by the present invention, the determining of the historical association data based on the values ​​of each non-diagonal element in the association matrix includes: Determining a total number of switching times based on the sum of the values ​​of each off-diagonal element in the association matrix; Based on the values ​​of the non-diagonal elements in the association matrix and the total number of switching times, an association ratio is determined, and based on the association ratio, the historical association data is determined.

[0012] According to a training method for a performance optimization model of a terminal device provided by the present invention, the historical application data is determined based on an array of usage records of the multiple applications; the array of usage records includes the usage interval duration of a single application, the number of times it is opened within the same day, the average time for each opening, and the time period for the use of a single application.

[0013] The present invention also provides a terminal device performance optimization device, comprising the following units: A first acquisition unit, used to acquire in real time the operation data of multiple applications installed in the terminal device; A scheduling strategy generating unit is used to input the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; A scheduling instruction generating unit, used to generate an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained by training based on a plurality of samples, wherein the samples include historical operation data of the application at a plurality of moments, and application control strategies for successive time periods associated with the historical operation data.

[0014] The present invention also provides a training device for a performance optimization model of a terminal device, comprising the following units: A second acquisition unit, configured to acquire historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; A third acquisition unit, configured to acquire an application control policy label based on a successive time period corresponding to the historical operation data; A training sample generating unit, configured to generate a training sample based on the historical operation data and an application control policy label corresponding to the historical operation data; A training unit is used to train the performance optimization model based on the training samples.

[0015] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a terminal device performance optimization method as described in any one of the above, or implements a training method for a performance optimization model of the terminal device as described in any one of the above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for optimizing the performance of a terminal device as described in any one of the above is implemented, or the method for training a performance optimization model of a terminal device as described in any one of the above is implemented.

[0017] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements a terminal device performance optimization method as described in any one of the above, or implements a training method for a performance optimization model of a terminal device as described in any one of the above.

[0018] The present invention provides a method, device and equipment for optimizing terminal device performance and training a performance optimization model for a terminal device. The performance optimization model of the method is obtained based on training of multiple samples. The samples include historical operation data of an application at multiple times, and application control policies for successive time periods associated with the historical operation data. During the application process, no manual design and maintenance by development engineers is required, which reduces development costs and workload, thereby compensating for the defects of the application scheduling policies with fixed priority sorting pre-designed in the prior art, improving the accuracy and flexibility of scheduling policy prediction, and further improving the accuracy and flexibility of application scheduling instruction generation, thereby further optimizing the performance of the terminal device and improving the user experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 It is a flow chart of the terminal equipment performance optimization method provided by the present invention.

[0021] Figure 2 It is a flowchart of step 121 in the terminal device performance optimization method provided by the present invention.

[0022] Figure 3 It is a flow chart of the scheduling strategy optimized based on user feedback operation provided by the present invention.

[0023] Figure 4 It is a flow chart of the training method of the performance optimization model of the terminal device provided by the present invention.

[0024] Figure 5 It is a flowchart of the steps of determining historical associated data provided by the present invention.

[0025] Figure 6 It is a flowchart of step 530 in the training method of the performance optimization model of the terminal device provided by the present invention.

[0026] Figure 7 It is a structural schematic diagram of a training device for a performance optimization model of a terminal device provided by the present invention.

[0027] Figure 8 It is a structural schematic diagram of the terminal equipment performance optimization device provided by the present invention.

[0028] Fig. 9 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Considering that existing intelligent systems often require manual design and maintenance by development engineers when optimizing terminal performance, this not only increases development costs and workload, but also the pre-designed fixed priority sorting may be difficult to meet users' personalized expectations and habits, thereby affecting user experience and acceptance, and reducing user experience and satisfaction.

[0031] Based on the above problems, the present invention simplifies the fixed priority sorting designed in advance by humans, reduces the systematic engineering operations of user data collection to development engineers and then feeding back to the code, thereby turning the "intelligent system" into a "true intelligent system".

[0032] The method provided by the present invention is described below from the performance optimization model training side and the performance optimization model application side.

[0033] Figure 1 is a flow chart of the terminal device performance optimization method provided by the present invention, such as Figure 1 As shown, the method includes step 110 , step 120 and step 130 .

[0034] Step 110, acquiring operation data of multiple applications installed in the terminal device in real time; Step 120, inputting the operation data into the performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; Step 130, generating an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained based on training of multiple samples, wherein the samples include historical operation data of the application at multiple times, and application control strategies for successive time periods associated with the historical operation data.

[0035] Specifically, first, the performance optimization model is trained, and then the model application is performed based on the performance optimization model after the performance optimization model training is completed. Here, the performance optimization model is obtained based on multiple sample training, wherein the samples include historical operation data of the application at multiple times, and the application control strategy for the successive time periods associated with the historical operation data.

[0036] Here, historical operation data is used to reflect the user's historical usage habits in the application, and the historical operation data may include historical application data and historical association data. Historical application data is used to reflect the user's historical usage of the application, and can be determined based on the usage record array of multiple applications, where the usage record array includes the usage interval length of a single application, the number of times it is opened within the same day, the average time of each opening, and the time period of use of a single application.

[0037] The historical association data is used to reflect the historical association between multiple applications. The historical association data can be the number of switching between multiple applications, the association ratio between multiple applications, or the specific association objects between multiple applications, etc. The embodiment of the present invention does not make specific limitations on this.

[0038] It should be noted that the application program may include APP, desktop application programs (Desktop Applications), network application programs (Web Applications) and embedded device programs, etc., and the embodiment of the present invention does not specifically limit this.

[0039] The subsequent time period refers to a time period in the future of the current moment, which may be the next moment of the current moment or the next time period of the current moment, and the embodiment of the present invention does not make any specific limitation to this.

[0040] For example, for applications A, B, and C, the historical running data of the applications are shown in Table 1: Table 1. Historical operation data of the application The usage interval in Table 1 is the interval between closing and reopening the application. The time period for using a single application is generally divided into four time periods: morning, noon, afternoon, evening, and midnight.

[0041] Historical correlation data can be the correlation of applications, for example, which applications are used together before and after using a certain application, and the order at the timeline level, that is, they are close to each other, for example, we use the previous three applications as a benchmark to judge the correlation. For applications that are opened repeatedly, it can be understood that within the usage cycle, the number of repeated calls between each two applications is sorted, and the more repeated calls, the stronger the correlation.

[0042] It can be understood that each row in the table can constitute an array of application usage records, for example: {(10, 2,360, 1, 30), (200, 30 , 5, 4, 20), … …}, which is interpreted as: 10 (usage interval duration), 2 (twice a day), 360 (6*60min), 1 (representing the morning), 30 (representing the number of consecutive switches), corresponding to Table 1.

[0043] Here, the application control policy refers to the control operation on the application based on historical operation data, which can be killing the process, adjusting the priority between applications, clearing the background of the application and preloading the application, etc. The embodiment of the present invention does not make specific limitations on this.

[0044] It should be noted that the performance optimization model can prioritize killing applications that are frequently used for more than one day in the historical application data when dealing with performance issues. For applications with short intervals and high usage rates, and applications that are frequently switched within the interval, that is, applications with strong correlation, the performance optimization model generates the predicted probability of the application based on the historical application data and historical correlation data, and in this process, it can remove noise with large deviations, thereby continuously optimizing the performance of the performance optimization model.

[0045] It can be understood that the performance optimization model is trained based on the historical operation data of multiple applications at multiple times, as well as samples of application control policies for successive time periods associated with the historical operation data, so that the performance optimization model has the ability to predict application control policies based on historical operation data.

[0046] When applying, first, the operation data of multiple applications installed in the terminal device is obtained in real time. Here, the operation data includes application data and related data. The terminal device can be a mobile phone, a car computer, etc., and the embodiment of the present invention does not make specific limitations on this.

[0047] Furthermore, the method provided in the embodiment of the present application can use the above-mentioned trained performance optimization model to input input data, for example, the real-time acquisition of operating data of multiple applications installed in the terminal device in the present invention, into the trained performance optimization model to obtain output data, for example, the scheduling strategy for the operation of the application in the present invention.

[0048] After the scheduling strategy is obtained, an application scheduling instruction corresponding to the scheduling strategy is generated, wherein the application scheduling instruction is used to optimize the performance of the terminal device. Here, the application scheduling instruction is an instruction covering the scheduling order of multiple applications.

[0049] That is, the performance optimization model can generate application scheduling instructions corresponding to the scheduling strategy, thereby optimizing the performance of the terminal device.

[0050] It should be noted that the performance optimization model can be a large language model (Large Language Model, LLM), and the scheduling strategy for predicting the operation of the application can be obtained by deploying each intelligent terminal device with a large language model. The large language model can be a pre-trained language model such as iFlytek Spark model (IFlytek Spark), XLNet model (extreme Multi-label Learning Network), ROBERTa model (Robustly Optimized BERT approach), etc. The embodiment of the present invention does not specifically limit this.

[0051] The large language model here is a deep learning algorithm that can perform various natural language processing (NLP) tasks. It uses multiple transformer models at the bottom, which are actually a group of neural networks. LLM is a very large deep learning model trained with massive data sets, which is also the powerful basis for it to recognize program interfaces, translate, predict or generate text or other content.

[0052] Specifically, program calls usually include function calls, which is an important function of LLM to connect external tools. This function enables LLM to interact seamlessly with external applications, significantly enhancing their capabilities. LLM can convert natural language into application calls corresponding to operable program interfaces, thereby changing the control strategy of the application.

[0053] It is understandable that the large language model has been able to predict the scheduling strategy of applications through the learning of historical operation data and the application control strategy of the successive time periods associated with the historical operation data. When we input the operation data, the large language model can deeply understand and form an effective reasoning chain. This series of processing enables the large language model to derive the scheduling strategy of the corresponding application in an accurate and efficient manner, thereby making up for the defects of the pre-designed fixed priority application scheduling strategy in the prior art, improving the accuracy and reliability of the application scheduling strategy, further optimizing the performance of the terminal device, and improving the user experience and satisfaction.

[0054] The method provided by the embodiment of the present invention first obtains the operation data of multiple applications installed in the terminal device in real time, then inputs the operation data into the performance optimization model, obtains the scheduling strategy for predicting the operation of the application, and finally generates the application scheduling instructions corresponding to the scheduling strategy, and the application scheduling instructions are used to optimize the performance of the terminal device. The performance optimization model of the method is obtained based on multiple sample training, and the samples include the historical operation data of the application at multiple times, and the application control strategy for the successive time periods associated with the historical operation data. During the application process, no manual design and maintenance by the development engineer is required, which reduces the development cost and workload, thereby compensating for the defects of the fixed priority application scheduling strategy pre-designed in the prior art, improving the accuracy and flexibility of the scheduling strategy prediction, and then improving the accuracy and flexibility of the application scheduling instruction generation, further optimizing the performance of the terminal device, and improving the user experience and satisfaction.

[0055] It should be noted that the model application method and the training method of the performance optimization model of the terminal device provided in the embodiment of the present invention are inventions based on the same concept, and can also be understood as two parts in a system, or two stages of an overall process: such as the model training stage and the model application stage.

[0056] Based on the above embodiment, the operation data includes application data and associated data, and step 120 includes: Step 121: input the operating data into the performance optimization model, generate the predicted probability of the application by the performance optimization model, and determine the scheduling strategy based on the predicted probability of the application.

[0057] Specifically, the operation data includes application data and related data, wherein the application data is used to reflect the real-time usage of the application by the user. The application data may also include the usage interval duration of a single application, the number of times it is opened within the same day, the average time for each opening, and the time period for the use of a single application.

[0058] The association data is used to reflect the real-time association status between multiple applications. The association data can be the number of switching times of multiple applications, the association ratio between multiple applications, or the specific association objects between multiple applications, etc. The embodiment of the present invention does not specifically limit this.

[0059] The running data can be input into the performance optimization model, and the performance optimization model generates the predicted probability of the application. Here, the performance optimization model can be a probability algorithm based on the underlying large language model, and the parameters of the model can be: in, represents the output of the linear layer, Represents weights, Indicates the running data of the application. represents the bias term, represents the sigmoid activation function, and A represents the predicted probability of the application.

[0060] It should be understood that after the performance optimization model is trained, its model parameters, such as and They are all determined, so the model can be applied based on the determined model parameters to generate the predicted probability of the application, and the scheduling strategy can be determined based on the predicted probability of the application. For example, the application with a larger predicted probability has an earlier corresponding scheduling strategy; for another example, the application with a smaller predicted probability has a later corresponding scheduling strategy.

[0061] Alternatively, the predicted probabilities of multiple applications may be sorted from large to small, and the scheduling strategies corresponding to the applications are executed in the order of the multiple applications in the sorting.

[0062] The method provided by the embodiment of the present invention inputs the operating data into the performance optimization model, generates the predicted probability of the application by the performance optimization model, and determines the scheduling strategy based on the predicted probability of the application, thereby more accurately determining the scheduling strategy of the application.

[0063] Based on the above embodiments, Figure 2 is a flow chart of step 121 in the terminal device performance optimization method provided by the present invention, such as Figure 2 As shown, the step 121 of determining the scheduling strategy based on the predicted probability of the application program includes: Step 1211, based on the performance optimization model, filter out a target prediction probability from the prediction probabilities of the application programs; Step 1212, determining a target application corresponding to the target prediction probability, and optimizing the scheduling strategy based on a program interface type corresponding to the target application.

[0064] Specifically, considering that there may be applications with smaller prediction probabilities in the output of the performance optimization model, it is necessary to adjust the process priority or kill the process for these applications to release the cache and avoid occupying too many resources. Therefore, based on the performance optimization model, the target prediction probability can be screened out from the prediction probabilities of the applications. Here, the target prediction probability is the smaller prediction probability among the prediction probabilities of multiple applications.

[0065] Then, the target application corresponding to the target prediction probability is determined, and the scheduling strategy is optimized based on the program interface type corresponding to the target application.

[0066] Specifically, determine the target application corresponding to the smaller target prediction probability, and then determine the program interface type corresponding to the target application. Here, the program interface type includes killing processes, adjusting priorities between applications, clearing the background of applications, etc. The embodiment of the present invention does not specifically limit this.

[0067] Here, we can determine which application to kill based on the program interface type corresponding to the target application, or if resources are sufficient, we can preload and pull up the application in the background to improve the experience. For example, the relationship between game application 1 and instant messaging application 2 may be that there is frequent switching, and they are two applications with strong correlation and high frequency and high intensity use. In comparison, the beauty tool application may only be used occasionally. It is unlikely to edit photos while playing games. After this logic is decided by the performance optimization model, it is executed by the performance optimization model and background management is performed, so as to close applications with low probability and preload applications with strong correlation, further improving the user experience and satisfaction.

[0068] It can be understood that during training, the performance optimization model learns to change the control strategy of the application by converting natural language into application calls corresponding to an operable program interface.

[0069] The method provided by the embodiment of the present invention, based on the performance optimization model, screens out the target prediction probability from the prediction probability of the application, and then determines the target application corresponding to the target prediction probability, and optimizes the scheduling strategy based on the program interface type corresponding to the target application, thereby making up for the defects of the application scheduling strategy with fixed priority sorting pre-designed in the prior art, and optimizing the scheduling strategy based on the program interface type corresponding to the target application, thereby improving the accuracy and flexibility of the scheduling strategy prediction, and further improving the accuracy and flexibility of the application scheduling instruction generation.

[0070] Based on the above embodiments, Figure 3 is a flow chart of optimizing the scheduling strategy based on user feedback operation provided by the present invention, such as Figure 3 As shown, step 130, before that, also includes: Step 131, in response to a user's feedback operation, comparing the feedback operation with the scheduling strategy to obtain a prediction accuracy rate and a number of incorrect predictions; the feedback operation includes manually launching an application and manually killing an application; Step 132: Optimize the scheduling strategy based on the prediction accuracy, the number of incorrect predictions and the operating data.

[0071] Specifically, considering that before generating the application scheduling instructions corresponding to the scheduling policy, the user may make some feedback operations, these feedback operations are used to feedback the user's habits. For example, when there are 3 applications loaded at the same time, but the user sometimes does not need to pull up these applications, the user may manually kill one or two of the applications.

[0072] Accordingly, in response to the user's feedback operation, the feedback operation and the scheduling strategy are compared to obtain the prediction accuracy and the number of wrong predictions, wherein the feedback operation includes manually launching the application and manually killing the application.

[0073] Then, based on the prediction accuracy, the number of incorrect predictions and the operating data, the scheduling strategy is optimized. Here, the operating data can be the usage time and opening time of the application at this time.

[0074] For example, for the three applications in the scheduling strategy predicted by the performance optimization model, the user opened one of the three applications, and did not open the other two applications. At this time, it can be determined that the prediction accuracy is 0.333%, the number of wrong predictions is 2, and the running data of the application at this time is recorded, such as the application usage time, opening time, etc. Based on this, the performance optimization model can optimize the scheduling strategy based on the prediction accuracy, the number of wrong predictions and the running data, so that the scheduling strategy meets the user's usage needs and improves user satisfaction.

[0075] In addition, when users always use only one application at a specific time and in a specific scenario and do not need to use other applications, we can establish a new pattern based on the pattern discovered by the performance optimization model to serve as a pattern matching module between applications and scheduling policies in the performance optimization model.

[0076] Based on any of the above embodiments, Figure 4 is a flow chart of a training method for a performance optimization model of a terminal device provided by the present invention, such as Figure 4 As shown, the method includes: Step 410, obtaining historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; Step 420, obtaining an application control policy tag based on the successive time period corresponding to each historical operation data; Step 430, generating a training sample based on the historical operation data and the application control policy label corresponding to the historical operation data; Step 440: training the performance optimization model based on the training samples.

[0077] Specifically, the model training method provided in the embodiment of the present invention can specifically apply data processing methods such as data training, machine learning, and deep learning to perform symbolic and formalized intelligent information modeling, extraction, preprocessing, and training on training samples, and finally obtain a trained performance optimization model.

[0078] First, historical operation data and a performance optimization model may be obtained, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications.

[0079] Here, the parameters of the performance optimization model may be preset or randomly generated, and the embodiment of the present invention does not specifically limit this.

[0080] Here, historical application data is used to reflect the user's historical usage of the application, which can be determined based on a usage record array of multiple applications. The usage record array includes the usage interval length of a single application, the number of times it is opened within the same day, the average time for each opening, and the time period for the use of a single application.

[0081] The historical association data is used to reflect the historical association between multiple applications. The historical association data can be the number of switching between multiple applications, the association ratio between multiple applications, or the specific association objects between multiple applications, etc. The embodiment of the present invention does not make specific limitations on this.

[0082] Then, the application control policy label based on the successive period corresponding to each historical operation data is obtained. Here, the successive period refers to a period in the future of the current moment, which can be the next moment of the current moment or the next period of the current moment, and the embodiment of the present invention does not specifically limit this.

[0083] Then, training samples are generated based on the historical operation data and the application control policy labels corresponding to the historical operation data.

[0084] Finally, the performance optimization model is trained based on the training samples. Specifically, the historical operation data is input into the performance optimization model to obtain the prediction scheduling strategy for predicting the operation of the application. After the prediction scheduling strategy is obtained based on the performance optimization model, the target loss can be calculated based on the difference between the prediction scheduling strategy and the application control strategy label, and the performance optimization model is iterated based on the target loss. The performance optimization model after the parameter iteration is used as the final trained performance optimization model.

[0085] It can be understood that the greater the difference between the predicted scheduling policy and the application control policy labels, the greater the target loss; the smaller the difference between the predicted scheduling policy and the application control policy labels, the smaller the target loss.

[0086] The method provided by the embodiment of the present invention obtains historical operation data, which includes historical application data for a single application and historical association data between multiple applications, and then obtains application control policy labels based on successive time periods corresponding to each historical operation data. Then, based on the historical operation data and the application control policy labels corresponding to the historical operation data, training samples are generated. Finally, based on the training samples, a performance optimization model is trained, so that the performance optimization model has the ability to accurately generate scheduling strategies for historical operation data, thereby making up for the defects of the pre-designed fixed priority application scheduling strategies in the prior art and improving the accuracy and flexibility of scheduling strategy predictions.

[0087] Based on any of the above embodiments, Figure 5 is a flow chart of the steps for determining historical associated data provided by the present invention, such as Figure 5 As shown, the step of determining the historical associated data includes: Step 510, obtaining the user's switching behavior data between the multiple applications; Step 520, constructing a correlation matrix based on the switching behavior data; the off-diagonal elements in the correlation matrix represent the number of switching between two applications, and the diagonal elements in the correlation matrix are zero; Step 530, determining the historical association data based on the values ​​of each non-diagonal element in the association matrix; The historical association data indicates the relative frequency of switching between two application programs associated with the element.

[0088] Specifically, the switching behavior data of the user between multiple applications may be obtained. Here, the switching behavior data may be the number of times the application A switches to the application B.

[0089] Then, based on the switching behavior data, a correlation matrix is ​​constructed, where the off-diagonal elements in the correlation matrix represent the number of switches between two applications, and the diagonal elements in the correlation matrix are zero.

[0090] Finally, based on the values ​​of each non-diagonal element in the association matrix, historical association data is determined, wherein the historical association data represents the relative frequency of switching between two applications associated with the element. Here, the values ​​of each non-diagonal element in the association matrix, that is, the number of switching between two applications, can be directly used as historical association data, or the values ​​of each non-diagonal element can be processed and used as historical association data, which is not specifically limited in the embodiment of the present invention.

[0091] For example, the currently used application is A, the previously used application is B, and the previously used application is C. Then the relative frequency of switching in the retention of these three applications is calculated and recorded in an N*N association matrix of, for example, N applications.

[0092] The current ranking of the above applications is: (ABC...), that is, the currently used application is ranked first, the previously used application is ranked after the currently used application, and so on...

[0093] For example, A switches with B 5 times, and then A switches with C 3 times (at this time B is ranked third, but has not been eliminated and is still being calculated). If A continues to switch with B 3 more times, a total of 8 times, and then A switches with D 1 time, when the number of times A and D switch exceeds the number of times A and BC switch each other, then the three most frequently switched ones will be retained, and the other one will be automatically eliminated.

[0094] It can be understood that the historical association data represents the relative frequency of switching between two applications associated with the element, that is, the historical association data is equivalent to the weight, and the value range of the weight value is 0-100%.

[0095] In addition, each application may be used as an anchor to calculate the switching frequency ranking, and the switching frequency ranking may be used as the historical association data, which is not specifically limited in the embodiment of the present invention.

[0096] Based on any of the above embodiments, Figure 6 is a flow chart of step 530 in the training method of the performance optimization model of the terminal device provided by the present invention, such as Figure 6 As shown, step 530 includes: Step 531, determining the total number of switching times based on the sum of the values ​​of each non-diagonal element in the association matrix; Step 532: determine a correlation ratio based on the values ​​of the non-diagonal elements in the correlation matrix and the total number of switching times, and determine the historical correlation data based on the correlation ratio.

[0097] Specifically, the total number of switching times may be determined based on the sum of the values ​​of each non-diagonal element in the association matrix.

[0098] Furthermore, the association ratio may be determined based on the value of each non-diagonal element in the association matrix and the total number of switching times. For example, the association ratio may be obtained by dividing the value of each non-object element by the total number of switching times.

[0099] Based on the association ratio, historical association data is determined. For example, the square of the association ratio can be used as the historical association data; for another example, the cube of the association ratio can be used as the historical association data, etc. This embodiment of the present invention does not specifically limit this.

[0100] Based on any of the above embodiments, a method for optimizing terminal device performance comprises the following steps: The first step is to obtain historical operation data and application control policy labels based on the successive time periods corresponding to each historical operation data. The historical operation data includes historical application data for a single application and historical association data between multiple applications. The historical application data is determined based on the usage record array of multiple applications. The usage record array includes the usage interval length of a single application, the number of times it is opened within the same day, the average time of each opening, and the time period of use of a single application.

[0101] The steps of determining the historical related data include: Step S1, obtaining the user's switching behavior data between the multiple applications; Step S2: construct an association matrix based on the switching behavior data; the off-diagonal elements in the association matrix represent the number of switching between two applications, and the diagonal elements in the association matrix are zero; Step S3: determining historical association data based on the values ​​of each non-diagonal element in the association matrix; the historical association data represents the relative frequency of switching between two application programs associated with the element.

[0102] Step S4, determining the total number of switching times based on the sum of the values ​​of each non-diagonal element in the association matrix; Step S5: Determine the association ratio based on the value of each non-diagonal element in the association matrix and the total number of switching times, and determine the historical association data based on the association ratio.

[0103] The second step is to generate training samples based on the historical operation data and the application control policy labels corresponding to the historical operation data.

[0104] The third step is to train the performance optimization model based on the training samples.

[0105] The fourth step is to obtain the operation data of multiple applications installed in the terminal device in real time, and the operation data includes application data and associated data.

[0106] In the fifth step, the operating data is input into the performance optimization model, and the performance optimization model generates the predicted probability of the application.

[0107] In the sixth step, based on the performance optimization model, the target prediction probability is screened out from the prediction probability of the application, and the target application corresponding to the target prediction probability is determined. Based on the program interface type corresponding to the target application, the scheduling strategy is optimized.

[0108] In the seventh step, in response to the user's feedback operation, the feedback operation and the scheduling strategy are compared to obtain the prediction accuracy and the number of incorrect predictions. The feedback operation includes manually launching the application and manually killing the application.

[0109] Step 8: Optimize the scheduling strategy based on prediction accuracy, number of incorrect predictions, and operating data.

[0110] The ninth step is to generate an application scheduling instruction corresponding to the scheduling strategy, where the application scheduling instruction is used to optimize the performance of the terminal device.

[0111] The method provided by the embodiment of the present invention, during the application process of the performance optimization model, utilizes the underlying probability matching logic of the performance optimization model to add the user's habit data (application data) and related data to the consideration scope of the performance optimization model, thereby making a comprehensive judgment to optimize the scheduling strategy.

[0112] The training device for the performance optimization model of the terminal device provided by the present invention is described below. The training device for the performance optimization model of the terminal device described below and the training method for the performance optimization model of the terminal device described above can refer to each other.

[0113] Based on any of the above embodiments, the present invention provides a training device for a performance optimization model of a terminal device. Figure 7is a schematic diagram of the structure of the training device of the performance optimization model of the terminal device provided by the present invention, such as Figure 7 As shown, the device comprises: A second acquisition unit 710 is used to acquire historical operation data, where the historical operation data includes historical application data for a single application and historical association data between multiple applications; A third acquisition unit 720 is used to acquire an application control policy label corresponding to a successive time period based on the historical operation data; A training sample generating unit 730 is used to generate a training sample based on the historical operation data and the application control policy label corresponding to the historical operation data; The training unit 740 is used to train the performance optimization model based on the training samples.

[0114] The device provided by the embodiment of the present invention obtains historical operation data, which includes historical application data for a single application and historical association data between multiple applications, and then obtains application control policy labels based on the successive time periods corresponding to each historical operation data. Then, based on the historical operation data and the application control policy labels corresponding to the historical operation data, training samples are generated. Finally, based on the training samples, a performance optimization model is trained, so that the performance optimization model has the ability to accurately generate scheduling strategies for the historical operation data, thereby making up for the defects of the fixed priority application scheduling strategies pre-designed in the prior art and improving the accuracy and flexibility of scheduling strategy predictions.

[0115] Based on any of the above embodiments, it further includes a unit for obtaining historical association data, and the unit for obtaining historical association data specifically includes: A switching behavior data acquisition unit is used to acquire the switching behavior data of the user between the plurality of applications; A correlation matrix construction unit is used to construct a correlation matrix based on the switching behavior data; the off-diagonal elements in the correlation matrix represent the number of switching between two applications, and the diagonal elements in the correlation matrix are zero; A historical association data acquisition subunit is used to determine the historical association data based on the value of each non-diagonal element in the association matrix; The historical association data indicates the relative frequency of switching between two application programs associated with the element.

[0116] Based on any of the above embodiments, the subunit for obtaining historical association data is specifically used for: Determining a total number of switching times based on the sum of the values ​​of each off-diagonal element in the association matrix; Based on the values ​​of the non-diagonal elements in the association matrix and the total number of switching times, an association ratio is determined, and based on the association ratio, the historical association data is determined.

[0117] Based on any of the above embodiments, the historical application data is determined based on the usage record array of the multiple applications; the usage record array includes the usage interval duration of a single application, the number of times it is opened within the same day, the average time for each opening, and the time period for the use of a single application.

[0118] The terminal device performance optimization device provided by the present invention is described below. The terminal device performance optimization device described below and the terminal device performance optimization method described above can be referenced to each other.

[0119] Based on any of the above embodiments, the present invention provides a terminal device performance optimization device, Figure 8 is a schematic diagram of the structure of the terminal equipment performance optimization device provided by the present invention, such as Figure 8 As shown, the device comprises: A first acquisition unit 810 is used to acquire in real time the operation data of multiple applications installed in the terminal device; A scheduling strategy generating unit 820 is used to input the operation data into the performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; A scheduling instruction generating unit 830 is used to generate an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained by training based on a plurality of samples, wherein the samples include historical operation data of the application at a plurality of moments, and application control strategies for successive time periods associated with the historical operation data.

[0120] The device provided by the embodiment of the present invention first obtains the operation data of multiple applications installed in the terminal device in real time, then inputs the operation data into the performance optimization model, obtains the scheduling strategy for predicting the operation of the application, and finally generates the application scheduling instructions corresponding to the scheduling strategy, and the application scheduling instructions are used to optimize the performance of the terminal device. The performance optimization model of this method is obtained based on multiple sample training, and the samples include the historical operation data of the application at multiple times, and the application control strategy for the successive time periods associated with the historical operation data. During the application process, no manual design and maintenance by the development engineer is required, which reduces the development cost and workload, thereby compensating for the defects of the fixed priority application scheduling strategy pre-designed in the prior art, improving the accuracy and flexibility of the scheduling strategy prediction, and then improving the accuracy and flexibility of the application scheduling instruction generation, further optimizing the performance of the terminal device, and improving the user experience and satisfaction.

[0121] Based on any of the above embodiments, the generating scheduling strategy unit 820 specifically includes: A scheduling strategy generation subunit is used to input the operating data into the performance optimization model, generate the predicted probability of the application by the performance optimization model, and determine the scheduling strategy based on the predicted probability of the application; the operating data includes application data and associated data.

[0122] Based on any of the above embodiments, the generating scheduling strategy subunit is specifically used for: Based on the performance optimization model, selecting a target prediction probability from the prediction probabilities of the application; A target application corresponding to the target prediction probability is determined, and the scheduling strategy is optimized based on a program interface type corresponding to the target application.

[0123] Based on any of the above embodiments, it further includes a feedback unit, wherein the feedback unit is specifically used to: In response to a user's feedback operation, the feedback operation is compared with the scheduling strategy to obtain a prediction accuracy rate and a number of incorrect predictions; the feedback operation includes manually launching an application and manually killing an application; The scheduling strategy is optimized based on the prediction accuracy, the number of incorrect predictions and the operating data.

[0124] Fig. 9 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Fig. 9As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930 and a communication bus 940, wherein the processor 910, the communication interface 920 and the memory 930 communicate with each other through the communication bus 940. The processor 910 may call the logic instructions in the memory 930 to execute the terminal device performance optimization method, the method comprising: obtaining the operation data of multiple applications installed in the terminal device in real time; inputting the operation data into the performance optimization model to obtain the scheduling strategy for predicting the operation of the application; generating the application scheduling instruction corresponding to the scheduling strategy, the application scheduling instruction is used to optimize the performance of the terminal device; wherein the performance optimization model is obtained based on multiple sample training, the samples include the historical operation data of the application at multiple times, and the application control strategy of the successive time period associated with the historical operation data.

[0125] The processor 910 can also call the logic instructions in the memory 930 to execute a training method for a performance optimization model of the terminal device, the method including: obtaining historical operation data, the historical operation data including historical application data for a single application and historical association data between multiple applications; obtaining an application control policy label based on a successive time period corresponding to each historical operation data; generating training samples based on the historical operation data and the application control policy label corresponding to the historical operation data; and training the performance optimization model based on the training samples.

[0126] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0127] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the terminal device performance optimization method provided by the above methods, and the method includes: acquiring operation data of multiple applications installed in the terminal device in real time; inputting the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application; generating an application scheduling instruction corresponding to the scheduling strategy, and the application scheduling instruction is used to optimize the performance of the terminal device; wherein the performance optimization model is obtained based on multiple sample training, and the samples include historical operation data of the application at multiple times, and application control strategies for subsequent time periods associated with the historical operation data.

[0128] When the computer program is executed by the processor, the computer can execute the training method of the performance optimization model of the terminal device provided by the above methods, which method includes: obtaining historical operation data, the historical operation data including historical application data for a single application and historical association data between multiple applications; obtaining an application control policy label based on the successive time period corresponding to each historical operation data; generating training samples based on the historical operation data and the application control policy label corresponding to the historical operation data; and training the performance optimization model based on the training samples.

[0129] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the terminal device performance optimization method provided by the above-mentioned methods, the method comprising: acquiring operation data of multiple applications installed in the terminal device in real time; inputting the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application; generating an application scheduling instruction corresponding to the scheduling strategy, the application scheduling instruction being used to optimize the performance of the terminal device; wherein the performance optimization model is obtained based on multiple sample training, the samples including historical operation data of the application at multiple times, and application control strategies for successive time periods associated with the historical operation data.

[0130] When the computer program is executed by the processor, it is implemented to execute the training method of the performance optimization model of the terminal device provided by the above-mentioned methods, and the method includes: obtaining historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; obtaining an application control policy label based on the successive time period corresponding to each historical operation data; generating training samples based on the historical operation data and the application control policy label corresponding to the historical operation data; and training the performance optimization model based on the training samples.

[0131] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0132] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing terminal device performance, characterized in that: include: Obtain real-time operation data of multiple applications installed in terminal devices; Inputting the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; generating an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained based on training of multiple samples, wherein the samples include historical operation data of the application at multiple times, and application control strategies for successive time periods associated with the historical operation data.

2. The terminal device performance optimization method according to claim 1, characterized in that: The operation data includes application data and associated data, and the inputting of the operation data into the performance optimization model to obtain a scheduling strategy for predicting the operation of the application program includes: The operation data is input into the performance optimization model, the performance optimization model generates the predicted probability of the application, and the scheduling strategy is determined based on the predicted probability of the application.

3. The terminal device performance optimization method according to claim 2, characterized in that: The determining the scheduling strategy based on the predicted probability of the application program includes: Based on the performance optimization model, selecting a target prediction probability from the prediction probabilities of the application; A target application corresponding to the target prediction probability is determined, and the scheduling strategy is optimized based on a program interface type corresponding to the target application.

4. The terminal device performance optimization method according to claim 1, characterized in that: The step of generating a corresponding application scheduling instruction based on the scheduling strategy also includes: In response to a user's feedback operation, the feedback operation is compared with the scheduling strategy to obtain a prediction accuracy rate and a number of incorrect predictions; the feedback operation includes manually launching an application and manually killing an application; The scheduling strategy is optimized based on the prediction accuracy, the number of incorrect predictions and the operating data.

5. A method for training a performance optimization model of a terminal device, characterized in that: include: Acquire historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; Obtaining an application control policy label for a successive period corresponding to each historical operation data; Generate a training sample based on the historical operation data and the application control policy label corresponding to the historical operation data; The performance optimization model is trained based on the training samples.

6. The method for training a performance optimization model of a terminal device according to claim 5, characterized in that: The step of determining the historical associated data includes: Acquire user switching behavior data between the multiple applications; Based on the switching behavior data, a correlation matrix is ​​constructed; the off-diagonal elements in the correlation matrix represent the number of switching between two applications, and the diagonal elements in the correlation matrix are zero; Determining the historical association data based on the values ​​of each off-diagonal element in the association matrix; The historical association data indicates the relative frequency of switching between two application programs associated with the element.

7. The method for training a performance optimization model of a terminal device according to claim 6, characterized in that: The determining the historical association data based on the values ​​of each non-diagonal element in the association matrix includes: Determining a total number of switching times based on the sum of the values ​​of each off-diagonal element in the association matrix; Based on the values ​​of the non-diagonal elements in the association matrix and the total number of switching times, an association ratio is determined, and based on the association ratio, the historical association data is determined.

8. The method for training a performance optimization model of a terminal device according to any one of claims 5 to 7, characterized in that: The historical application data is determined based on a usage record array of the multiple applications; the usage record array includes the usage interval duration of a single application, the number of times it is opened within a day, the average time of each opening, and the time period of use of a single application.

9. A terminal equipment performance optimization device, characterized in that: include: A first acquisition unit, used to acquire in real time the operation data of multiple applications installed in the terminal device; A scheduling strategy generating unit is used to input the operation data into a performance optimization model to obtain a scheduling strategy for predicting the operation of the application program; A scheduling instruction generating unit, used to generate an application scheduling instruction corresponding to the scheduling policy, wherein the application scheduling instruction is used to optimize the performance of the terminal device; The performance optimization model is obtained by training based on a plurality of samples, wherein the samples include historical operation data of the application at a plurality of moments, and application control strategies for successive time periods associated with the historical operation data.

10. A training device for a performance optimization model of a terminal device, characterized in that: include: A second acquisition unit, configured to acquire historical operation data, wherein the historical operation data includes historical application data for a single application and historical association data between multiple applications; A third acquisition unit, configured to acquire an application control policy label based on a successive time period corresponding to the historical operation data; A training sample generating unit, configured to generate a training sample based on the historical operation data and an application control policy label corresponding to the historical operation data; A training unit is used to train the performance optimization model based on the training samples.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the terminal device performance optimization method according to any one of claims 1 to 4 is implemented; Or the processor implements the training method of the performance optimization model of the terminal device as described in any one of claims 5 to 8 when executing the computer program.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the terminal device performance optimization method as described in any one of claims 1 to 4; or when the computer program is executed by the processor, it implements the training method of the performance optimization model of the terminal device as described in any one of claims 5 to 8.

Images