Application program performance optimization method and device, equipment, medium and product

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an application program performance optimization method and device, equipment, a medium and a product. The method comprises the steps that source codes of an application program are obtained, and a plurality of optimization strategies are determined; obtaining a static optimization factor based on the optimization effect of the optimization strategy; analyzing the running log of the application program to obtain corresponding system index information; comparing the system index information with a preset index to obtain a dynamic optimization factor; utilizing a deep learning model to confirm a performance score vector of the application program; and determining a performance optimization scheme based on the performance score vector. The static performance of the application program can be analyzed in real time by performing exception analysis on the source code; by analyzing the running log during running of the application program, the suspicious behavior during running of the application program and the influence degree on the performance can be analyzed, the comprehensive performance conditions of program code development and real-time running are considered, and the method is more flexible and intelligent.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, and in particular, to an application program performance optimization method, device, equipment, medium and product. BACKGROUND

[0002] With the continuous development of mobile Internet, smart terminals equipped with various application programs have been popularized in all aspects of life. At the same time, users have increasingly high expectations for the software types and functions of application programs, especially Android application programs. In order to cope with this challenge, developers need to continuously optimize the performance of application programs to improve the usage and retention rate of application programs.

[0003] At present, the performance optimization of related technologies for application programs is mainly from the perspective of software development, without considering the problems that may occur in the real-time running of application programs, resulting in that this performance optimization method is not flexible and intelligent enough. In recent years, there have been some optimization methods for supervising and judging the performance of real-time running, but this method is not mature enough, has the disadvantages of insignificant results and large resource occupation, and is likely to affect the performance of the application program itself. SUMMARY

[0004] The present disclosure is proposed in view of the above problems, and provides an application program performance optimization method, device, equipment, medium and product.

[0005] According to one aspect of the present disclosure, an application program performance optimization method is provided, comprising: obtaining the source code of the application program, performing abnormality analysis on the source code, and determining a plurality of optimization strategies; obtaining a static optimization factor of the application program based on the optimization effect of the optimization strategy; obtaining the running log of the application program, performing analysis on the running log, obtaining system index information corresponding to the behavior sequence, wherein the behavior sequence includes a plurality of behaviors affecting the performance of the application program, and the index information represents the performance state of the application program when the behavior sequence occurs; comparing the system index information with the preset index to obtain a dynamic optimization factor of the application program; analyzing the static optimization factor and the dynamic optimization factor by using a pre-trained deep learning model to confirm a performance score vector of the application program, wherein the deep learning model is trained based on historical static optimization factors and historical dynamic optimization factors of known performance problems; confirming a performance optimization scheme of the application program based on the performance score vector.

[0006] The beneficial effects of the scheme are that the static optimization factor is confirmed by performing the exception analysis on the source code, the static performance of the application program can be analyzed in real time, the suspicious behavior of the application program at runtime and the influence degree on the performance can be analyzed by confirming the dynamic optimization factor by analyzing the running log of the application program at runtime. The scheme considers the all-round performance of the program code development and real-time running, can identify more performance problems, is more flexible and intelligent, and only needs the system index information when the behavior sequence occurs, and occupies less resource amount.

[0007] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the static optimization factor of the application program is obtained based on the optimization effect of the optimization strategy, including: The optimization coefficient of the optimization strategy is calculated based on the optimization effect of the optimization strategy; The optimization coefficients are sorted, and the largest n target optimization coefficients are screened; The static optimization factor of the application program is obtained based on the n target optimization coefficients.

[0008] In one or more embodiments, the beneficial effects of the scheme are that the optimization coefficients of various optimization strategies are calculated based on the performance before and after the optimization of various optimization strategies, so as to quantize the optimization strategies into calculable static optimization factors, which is beneficial to the performance optimization of the application program.

[0009] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the system index information is compared with the preset index, and the dynamic optimization factor of the application program is obtained, including: The system index information is compared with the preset index, and the deviation degree value of each system index information is obtained; The behavior sequence is compared with the pre-established performance problem feature library, and the coincidence degree of the behavior sequence and the problem behavior in the performance problem feature library is confirmed to reach a preset threshold; The dynamic optimization factor of the application program is calculated based on the deviation degree value of each system index information.

[0010] In one or more embodiments, the beneficial effects of the scheme are that by comparing the system index information with the preset index, it can be known which index of the application program is abnormal or deviated, for example, the CPU usage rate abnormally rises, the memory occupation grows too fast, etc.; by comparing the behavior sequence with the pre-established performance problem feature library, it can be known whether the behavior sequence is a problem behavior, for example, it is found that the behavior sequence is highly similar to the "suspicious jump behavior", so the behavior sequence is likely to be the "suspicious jump behavior". Therefore, according to the deviation degree value and the possibility of the behavior sequence being a problem behavior, the dynamic optimization factor can be generated, indicating that the application program has serious performance problems.

[0011] Further, according to the application performance optimization method of one aspect of the present disclosure, the process of establishing the performance problem feature library comprises: obtaining a historical behavior sequence of the application and historical system index information corresponding to the historical behavior sequence; taking a key behavior in the historical behavior sequence as a sequence point, recursively classifying the historical behavior sequence to obtain a plurality of sub-sequences, wherein the key behavior is a behavior that has a deep impact on the performance of the application, and each sub-sequence comprises a sequence; dividing the plurality of sub-sequences into an impact sequence and a normal sequence; obtaining a target sequence in the normal sequence that affects the performance of the application, and adding the target sequence to the impact sequence; generating the performance problem feature library based on m target impact sequences in the impact sequence that have the greatest impact on the performance of the application.

[0012] In one or more embodiments, the present scheme has the beneficial effect that: by taking a key behavior in the historical behavior sequence as a sequence point, recursively classifying the historical behavior sequence, and cutting the continuous behavior sequence into a structured behavior segment centered on the key behavior, it is helpful to accurately analyze which specific behavior triggers the performance problem. For example, the historical behavior sequence is “go home → open App → apply for location permission → render map → request network data →...”, and “apply for location permission” and “render map” are key behaviors. Then the historical behavior sequence is disassembled into behavior segments centered on “apply for location permission” and “render map”, such as a network request and an interface rendering. By bidirectional screening of the impact sequence and the normal sequence, the target sequence (i.e. the impact sequence) that is disguised or hidden can be screened out, and the target impact sequence with a small impact degree can be removed, and only the impact sequence with a large impact degree is retained, to generate the performance problem feature library. The performance problem feature library can clearly reveal which behavior or behavior combination is most likely to affect the performance of the application. When the real-time behavior sequence is obtained, it can be directly compared with the performance problem feature library, which is conducive to quickly identifying whether the behavior sequence is a problem behavior and improving the calculation efficiency.

[0013] Further, according to the application performance optimization method of one aspect of the present disclosure, the pre-trained deep learning model is used to analyze the static optimization factor and the dynamic optimization factor to confirm the performance score vector of the application, comprising: weighting and fusing the static optimization factor and the dynamic optimization factor to obtain a comprehensive feature factor; inputting the comprehensive feature factor into the pre-trained deep learning model to obtain a performance score vector output by the deep learning model, wherein the performance score vector is used to represent the similarity between the performance state of the application and the known performance problem.

[0014] In one or more embodiments, the beneficial effects of the present solution are that, by using the pre-trained deep learning model to analyze the comprehensive feature factors, the similarity between the comprehensive feature factors and various known performance problems can be quickly calculated, and the greatest similarity may be the current performance problem of the application program.

[0015] In addition, according to the application program performance optimization method of one aspect of the present disclosure, based on the performance score vector, the performance optimization scheme of the application program is confirmed, including: Based on the performance score vector, the severity of the influence of the current performance problem on the application program is confirmed; According to the severity, the performance optimization scheme of the application program is confirmed.

[0016] In one or more embodiments, the beneficial effects of the present solution are that, by analyzing the severity of the influence of the current performance problem on the application program, it can be determined which performance optimization scheme to execute, thereby achieving precise optimization of the application program.

[0017] In addition, according to the application program performance optimization method of one aspect of the present disclosure, after obtaining the source code of the application program, performing abnormality analysis on the source code, and determining a plurality of optimization strategies, the method further includes: Based on the optimization strategies, an optimization strategy matrix is constructed; wherein the number of elements included in the optimization strategy matrix is determined based on the number of optimization strategies, and the optimization strategy matrix is used to represent the optimization strategies included in the application program.

[0018] In one or more embodiments, the beneficial effects of the present solution are that, by constructing the optimization strategy matrix, it can be directly known which optimization strategies are included in the application program.

[0019] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the system index information is compared with the preset index to obtain the deviation degree value of each system index information, including: Quantifying the system index information into system index data; Inputting the system index data into a pre-established dynamic characteristic model, wherein the dynamic characteristic model includes historical system index information, and the dynamic characteristic model generates the preset index according to the historical system index information; Obtaining the deviation degree value of each system index information through the dynamic characteristic model.

[0020] In one or more embodiments, the beneficial effects of the present solution are that, during the running of the application program, running logs are continuously generated, and these running logs include various system index information, and by storing and analyzing the system index data through the dynamic characteristic model, more reasonable preset indexes can be calculated.

[0021] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the severity of the current performance problem on the application program is determined based on the performance score vector, including: When the performance score vector reaches a first threshold, the severity of the current performance problem on the application program is determined to be mild; When the performance score vector reaches a second threshold, the severity of the current performance problem on the application program is determined to be moderate; When the performance score vector reaches a third threshold, the severity of the current performance problem on the application program is determined to be severe; wherein the first threshold, the second threshold and the third threshold are increasing.

[0022] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the optimization coefficient of the optimization strategy is calculated based on the optimization effect of the optimization strategy, including: Obtaining initial performance data of the application program, wherein the initial performance data represents the performance of the application program before the optimization strategy is implemented; Obtaining optimized performance data of the application program, wherein the optimized performance data represents the performance of the application program after the optimization strategy is implemented; Calculating the ratio of the mean values of the initial performance data and the optimized performance data to obtain the optimization coefficient of the optimization strategy.

[0023] In addition, according to the application program performance optimization method of one aspect of the present disclosure, the precision and recall are set as evaluation parameters of the deep learning model; The precision is used to evaluate the probability that the comparison result output by the deep learning model is captured in the test set, and the recall is used to evaluate the proportion of the comparison result in the actual sample.

[0024] According to another aspect of the present disclosure, an application program performance optimization device is provided, including: An analysis module for obtaining source code of an application program, performing abnormal analysis on the source code, and determining a plurality of optimization strategies; A static optimization factor generation module for obtaining a static optimization factor of the application program based on the optimization effect of the optimization strategy; A running monitoring module for obtaining running logs of the application program when the application program is running, analyzing the running logs, and obtaining system index information corresponding to a behavior sequence, wherein the behavior sequence includes a plurality of behaviors affecting the performance of the application program, and the index information represents the performance state of the application program when the behavior sequence occurs; A dynamic optimization factor generation module for comparing the system index information with a preset index to obtain a dynamic optimization factor of the application program; The similarity calculation module is configured to analyze the static optimization factor and the dynamic optimization factor by using a pre-trained deep learning model to determine a performance score vector of the application program, wherein the deep learning model is trained based on historical static optimization factors and historical dynamic optimization factors of known performance problems. The optimization module is configured to determine a performance optimization scheme of the application program based on the performance score vector.

[0025] According to yet another aspect of the present disclosure, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to implement the method of the above aspect.

[0026] According to yet another aspect of the present disclosure, a computer readable storage medium is provided, which stores a computer program, the computer program being executed by a processor to implement the method of the above aspect.

[0027] According to yet another aspect of the present disclosure, a computer program product is provided, which includes a computer program, the computer program being executed by a processor to implement the method of the above aspect.

[0028] It is to be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further explanation of the subject technology. BRIEF DESCRIPTION OF DRAWINGS

[0029] The above and other objects, features and advantages of the present disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which like reference characters refer to the like elements throughout. The attached drawings are intended to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the present disclosure and serve to explain the present disclosure, and do not limit the present disclosure. In the drawings, like reference numerals refer to the same or similar elements throughout.

[0030] Figure 1 is a system architecture diagram illustrating application of an application program performance optimization method according to embodiments of the present disclosure.

[0031] Figure 2 is a flowchart illustrating generation of a static optimization factor according to embodiments of the present disclosure.

[0032] Figure 3 is a flowchart illustrating an application program performance optimization method according to embodiments of the present disclosure.

[0033] Figure 4 is a structural schematic diagram of an application program performance optimization apparatus according to embodiments of the present disclosure.

[0034] Figure 5 is a structural schematic diagram of a computer device according to embodiments of the present disclosure.

[0035] Figure 6 is a schematic diagram illustrating a computer program product according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0036] In order to make the purposes, technical solutions and advantages of the present disclosure more obvious, the example embodiments according to the present disclosure will be described in detail below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited to the example embodiments described herein.

[0037] With the continuous development of mobile Internet, smart terminals carrying various application programs have been popularized in all aspects of life. At the same time, users have higher and higher expectations for the software types and functions of application programs, especially Android application programs. In order to cope with this challenge, developers need to continuously optimize the performance of application programs to improve the usage and retention rate of users.

[0038] At present, the performance optimization of Android application programs in the related art is mainly from the perspective of software development, without considering the problems that may occur during the real-time running of the application program, resulting in that this performance optimization method is not flexible and intelligent enough. In recent years, there have been some optimization methods for supervising and judging the performance during real-time running, but this method is not mature enough, has the disadvantages of insignificant achievements and large resource occupation, and is likely to affect the performance of the application program itself.

[0039] The above describes an application program performance optimization method, device, equipment, medium and product according to an embodiment of the present disclosure with reference to the drawings, which can analyze the static performance of the application program in real time by confirming the static optimization factor through abnormal analysis of the source code; and can analyze the suspicious behavior of the application program during running and the influence degree on the performance by confirming the dynamic optimization factor through parsing the running log during running of the application program. The present solution considers the overall performance of the program code development and real-time running, can identify more performance problems, is more flexible and intelligent, and only needs system index information when the behavior sequence occurs, and occupies less resources.

[0040] For the convenience of understanding the present embodiment, firstly, a kind of application performance optimization method disclosed by the present embodiment is introduced in detail, the execution subject of the application performance optimization method provided by the present embodiment is generally computer equipment with certain computing power, which includes, for example: terminal equipment or server or other processing equipment, terminal equipment can be user equipment (User Equipment, UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital processing (Personal DiTital Assistant, PDA), handheld device, computing device, vehicle-mounted device, wearable device, etc.In some possible implementation ways, the application performance optimization method can be realized by the way that processor calls computer readable instructions stored in memory.

[0041] As shown in Figure 1 The system architecture diagram of the application performance optimization method provided by the present embodiment includes static optimization module 1, dynamic optimization module 2 and deep learning module 3, wherein the present embodiment takes the application program of Android system as optimization object to describe in detail, is applicable to cloud computer and local computer, and in actual application, adaptive modification can be carried out according to different operating systems (such as Android or IOS system, etc.), and the specific description of each module is as follows: Static optimization module 1: used for generating static optimization factor, as shown in Figure 2 The generation flow chart of static optimization factor includes S201-S204: S201: static code is analyzed and optimized.

[0042] Specifically, the static code (source code) of application program is analyzed and processed, and the code that affects the memory of application program is processed. The focus of abnormal analysis can be determined according to actual demand or actual running condition of application program, for the convenience of understanding, the present embodiment provides the following three kinds of abnormal analysis focus: (1) onCreate () startup speed optimization.

[0043] onCreate () method is the first method started when opening an Activity, and its function is to complete the necessary initialization work of starting an Activity, such as setting the layout resource of page, initializing component information, etc.If there are too many codes in onCreate (), the time consumption of this link is too long, which will cause the startup speed of application program to slow down. The present embodiment rewrites a private method initView (), and places all the statements of initializing various component information in this private method, and only executes initView () method in onCreate (), so as to solve the problem of too much code during startup.

[0044] (2) onDraw() drawing and rendering speed optimization.

[0045] Rendering refers to drawing various layout controls to the screen, and rendering speed represents the speed at which a complete page is presented to the user, so if you want to give the user a smooth user experience, you need to speed up the rendering speed. Among them, overdraw is that a pixel on the screen is drawn too many times in the same frame time, which is usually caused by too many layout control nesting or overlapping background colors.

[0046] The optimization process of onDraw() drawing is as follows: first, cancel the multi-layer nesting between containers and change it to a single-layer structure or two-layer nesting, then only set the white background to the outermost container, and cancel the background setting of the inner container and the control, in this way, multiple background colors can be avoided. Finally, delete some controls that have no output content, such as empty text boxes that serve as partition areas, etc., which can simplify the layout code.

[0047] (3) Response speed and memory optimization.

[0048] The main technical means or consideration angles for response speed and memory optimization are: 1) Reasonably use 4 Java references to optimize memory usage. Java provides 4 reference types: strong reference, soft reference, weak reference, and virtual reference, and selects the appropriate reference type according to different scenarios.

[0049] 2) Reduce unnecessary memory overhead. Reasonably use automatic boxing, use system resources, object pools, view reuse, and Bitmap reuse.

[0050] 3) ArrayMap can be used to replace enumeration types, and Lrucache cache can be used reasonably.

[0051] 4) Try to reduce system GC operations to increase application smoothness and reduce stuttering.

[0052] It should be noted that the focus of the above three exception analyses is only to illustrate by example, and is not limited to analyzing only these three points. In addition, for each of the above focuses, multiple optimization strategies can be designed, such as for "response speed and memory optimization", four optimization strategies can be designed, and the number is not limited.

[0053] S202: Construct an optimization strategy matrix.

[0054] Various optimization strategies are listed as a set , where k represents the number of optimization strategies, and the optimization strategy matrix is constructed as follows:

[0055] The optimization strategy matrix is represented by Gn x k, the number of strategies is n, representing the application whether to include the optimization strategy , if yes, the value is 1, otherwise the value is 0. Through the optimization strategy matrix, it can be directly known which optimization strategies are included in the application.

[0056] S203: Calculate the optimization coefficient of the optimization strategy.

[0057] Different optimization strategies will produce different optimization effects. In order to facilitate comparison, different optimization coefficients are set for different optimization strategies in this embodiment. The optimization coefficient is obtained from the ratio of the effect before optimization and the effect after optimization. Specifically, the calculation formula of the optimization coefficient is as follows:

[0058] wherein, represents the optimization coefficient of the optimization strategy , the set of performance before optimization is , and the mean of performance after optimization is

[0059] S204: Extract static optimization factors.

[0060] Specifically, the optimization coefficients calculated above are sorted. This embodiment adopts common Top-n sorting screening to extract the top n optimization coefficients to obtain , and is taken as a static optimization factor waiting for training.

[0061] The static optimization module 1 first performs static code analysis on the cloud computer application to be detected, and creatively integrates various technical means and analysis methods to comprehensively consider from multiple angles such as startup speed, rendering speed, response speed, etc. to calculate the static optimization factor.

[0062] Dynamic optimization module 2: for generating dynamic optimization factors, specifically including the following steps: Step 1: Monitor the dynamic performance of the application.

[0063] Among them, monitoring the dynamic performance includes: 1) Tool preparation: ensure that the necessary log grabbing tools are installed, such as adb (for Android), logcat (Android log tool), etc.

[0064] 2) Log grabbing: use the corresponding tool to grab the log file when the application is running, and ensure that all necessary system index information is included.

[0065] ​3) Log analysis: Write or use existing scripting tools to parse the running log, extract key system indicator information, such as CPU usage, memory consumption, packet transmission, wireless connection status, running process number, battery level, network transmission data volume, file read / write path and size, etc.

[0066] 4) Establish a dynamic behavior record table: Record the performance-affecting operations and key behaviors during application running, generate a running log containing timestamp, behavior module / component, behavior sequence, and real-time system indicator information.

[0067] Among them, the behavior sequence includes a series of behaviors sorted by time, such as: [start application → apply for location permission → render map → request network data → …], and the system indicator information corresponding to the behavior sequence refers to the performance of the application when the behavior sequence occurs.

[0068] Step 2: Compare the behavior sequence with the pre-established performance problem feature library.

[0069] Among them, the establishment process of the performance problem feature library includes S1-S5: S1: Obtain the historical behavior sequence of the application and the historical system indicator information corresponding to the historical behavior sequence to form a sample library.

[0070] S2: Recursively classify the historical behavior sequence by taking the key behaviors in the historical behavior sequence as sequence points, and obtain multiple sub-sequences.

[0071] Among them, the key behavior is a behavior that has a deep impact on the performance of the application, such as permission application, graphics rendering, network request, page jump, memory processing, etc. For example, the historical behavior sequence [start application → apply for location permission → render map → request network data → …], assuming that "apply for location permission", "render map", "request network data" are key behaviors, the historical behavior sequence is cut into 3 structured behavior fragments centered on key behaviors (such as a network request, an interface rendering) and classified, which is recursive classification. Recursive classification is to continuously subdivide until it cannot be further divided, ensuring that each sub-sequence only describes one core action.

[0072] S3: Divide the multiple sub-sequences into impact sequences and normal sequences.

[0073] Among them, the normal sequence includes surface normal behaviors, and the impact sequence includes "problem behaviors" that will affect the performance of the application.

[0074] S4: Obtain a target sequence in the normal sequence that affects the performance of the application program, and add the target sequence to the impact sequence.

[0075] The normal sequence may include some "problem behaviors" that are not easily detected and have a high degree of concealment. Therefore, the embodiment can expand the scope of investigation and avoid omissions by obtaining a target sequence (i.e., an impact sequence including "problem behaviors") in the normal sequence through more refined pattern recognition (such as slight abnormalities combined with system indicator information).

[0076] S5: Generate a performance problem feature library based on m target impact sequences in the impact sequence that have the greatest impact on the performance of the application program.

[0077] Specifically, the "problem behaviors" included in the impact sequence may have different degrees of impact on the performance of the application program. Therefore, behaviors that have some problems but have little actual impact are removed, and only m target impact sequences with the greatest impact are retained, which is conducive to focusing on behaviors that can seriously affect the performance of the application program, such as behaviors that cause lag, power consumption, etc.

[0078] It can be seen that the performance problem feature library records a large number of verified "problem behavior sequences" that have a significant impact on performance. Therefore, comparing the behavior sequence with the pre-established performance problem feature library can quickly identify whether the current behavior sequence is similar to some problem behaviors in the performance problem feature library (such as reaching a preset threshold of coincidence degree), to determine whether the application program has problems during runtime and which aspect of the problem.

[0079] Step 3: Analyze the dynamic performance of the application program.

[0080] Read the running log of the application program, extract the system indicator information therein, and include the corresponding system indicator information in the pre-constructed dynamic feature model. The dynamic feature model includes a large amount of historical system indicator information, and the dynamic feature model is used to set a preset indicator based on historical data, system specifications or business logic.

[0081] Specifically, the system indicator information is quantified into specific numerical values or states (i.e., system indicator data), such as CPU usage represented by a percentage, memory consumption represented by MB, data packet transmission represented by data size, etc. The quantified system indicator information is input into the dynamic feature model, and the dynamic feature model compares and analyzes the real-time system indicator information of the application program with the preset indicator to check whether the current system running state is consistent with the expected state, i.e., whether there is a deviation or anomaly, and finally obtains a deviation degree value of each system indicator information.

[0082] By comparing the system index information with the preset index, it can be known which index of the application program is abnormal or deviated, for example, abnormal surge of CPU usage rate, too fast growth of memory occupation, etc. By comparing the behavior sequence with the pre-established performance problem feature library, it can be known whether the behavior sequence is a problem behavior, for example, it is found that the behavior sequence is highly similar to the "suspicious jump behavior", so the behavior sequence is likely to be the "suspicious jump behavior". Therefore, according to the deviation degree value and the possibility of the behavior sequence being a problem behavior, a dynamic optimization factor can be generated, indicating that the application program has a serious performance problem.

[0083] Step 4: Extract the dynamic optimization factor.

[0084] Specifically, the deviation degree values of each system index information are combined into a feature vector as a dynamic optimization factor.

[0085] The dynamic optimization module 2 tracks and monitors the application program during running. The module first records the behavior of the application program during running to establish a log information table, and secondly establishes a common performance problem feature library affecting the application, and compares the behavior of the application program during running with the performance problem feature library for feature analysis and extraction. This embodiment considers the change of the system index information of the device during running of the application program, analyzes the situation causing abnormal fluctuation of the index information by continuously reading the key performance index information of the device, and comprehensively obtains the dynamic feature factor.

[0086] The deep learning module 3 is used for analyzing and comparing the similarity of the static optimization factor and the dynamic optimization factor. Specifically, the following process is included: Step 1: Weighted fusion of the static optimization factor and the dynamic optimization factor to obtain a comprehensive feature factor.

[0087] The calculation formula of the comprehensive feature factor is:

[0088] Among them, represents the comprehensive feature factor, represents the static optimization factor, represents the dynamic optimization factor, represents the weight coefficient.

[0089] Step 2: Input the comprehensive feature factor into the pre-trained deep learning model to obtain a performance score vector output by the deep learning model, wherein the performance score vector is used to represent the similarity between the performance state of the application program and the known performance problem.

[0090] Step 3: Based on the performance score vector, confirm the severity of the influence of the current performance problem on the application program.

[0091] For example, when the similarity is 0.7 or more, it is confirmed that the current performance problem has a slight impact on the application program (there may be some impact), when the similarity is 0.8 or more, it is confirmed that the impact is moderate (there is an impact), and when the similarity is close to 0.9, it is confirmed that the impact is severe (serious impact).

[0092] Step 4: According to the severity, the performance optimization scheme of the application program is confirmed.

[0093] The training process of the deep learning model includes: Model construction: The static optimization factor and the dynamic optimization factor of the static optimization module 1 and the dynamic optimization module 2 historical output are weighted and fused to calculate the corresponding comprehensive feature factor.

[0094] The calculation formula of the comprehensive feature factor is:

[0095] Among them, denotes the comprehensive feature factor, denotes the static optimization factor, denotes the dynamic optimization factor, denotes the weight coefficient.

[0096] Model training: The input contains the shallow features of the application program containing noise, which can be an Android application package (APK). The input data is trained layer by layer in the deep learning encoder model. The training formula is set as follows in this embodiment:

[0097]

[0098] Among them, denotes the number of current hidden layers, belongs to {1,...n}, W denotes the weight matrix, and b denotes the bias vector. denotes the fusion vector (containing noise) of the APK. After unsupervised learning layer by layer, the deeper APK implicit feature vector is output after decoding , which is the information source required by the model, which can be reconstructed into an APK basic vector , wherein belongs to , n is the total number of all behavior characteristics, and each behavior characteristic needs a corresponding neuron to process, is the total evaluation vector of the to-be-tested behavior. is the encoding function, for decoding function, representing the behavior characteristic parameter matrix (W, b).

[0099] Comparing the training results: the similarity of the model output characteristic factors, and the similarity calculation formula is:

[0100] wherein and respectively represent the new mixed factor characteristic vectors obtained by model training, and are respectively calculated by the behavior u and the behavior v through The similarity obtained by the embodiment in combination with the latest similarity calculation method is the similarity of the characteristic factors of the behavior and the mixed characteristic factors of the behavior to be detected.

[0101] Specifically, the embodiment sets precision (Precision) and recall (Recall) as two standard evaluation parameters for comparison, in order to facilitate the precision being marked as and the recall being marked as Both of these two indexes are used to measure the accuracy of the comparison result, but they are different in emphasis, the precision calculates the probability of the comparison result being captured in the test set, and the higher the value, the higher the accuracy of the comparison, the recall calculates the proportion of the comparison result set in the entire actual sample, and the higher the value, the wider the coverage of the algorithm. The calculation formula is as follows:

[0102]

[0103] wherein, is the result set of all comparison results, is the item set selected by the user on the test set, both of which take the intersection of the entire comparison result set and the selected item set on the test set as the numerator, and the difference is that the denominator is different, the denominator of the precision is the set of all user prediction results, and the denominator of the recall is the set of all selected test sets by the user.

[0104] The embodiment comprehensively selects the precision and the recall as the test parameters, and when the precision of the model training reaches 90% or more, the current model parameters can be used for comprehensive characteristic factor comparison calculation.

[0105] On the basis of the above embodiment, the embodiment further provides an application program performance optimization method, as shown in Figure 3 FIG. 1 is a flowchart of the application program performance optimization method, which comprises S301-S306: S301: Obtain the source code of the application program, perform exception analysis on the source code, and determine a plurality of optimization strategies.

[0106] S302: Obtain a static optimization factor of the application program based on an optimization effect of the optimization strategy.

[0107] S303: Obtain a running log of the application program during running of the application program, analyze the running log, and obtain system index information corresponding to a behavior sequence.

[0108] The behavior sequence includes a plurality of behaviors affecting the performance of the application program, and the index information represents the performance state of the application program when the behavior sequence occurs.

[0109] S304: Compare the system index information with a preset index, and obtain a dynamic optimization factor of the application program.

[0110] S305: Analyze the static optimization factor and the dynamic optimization factor by using a pre-trained deep learning model, and confirm a performance score vector of the application program.

[0111] The deep learning model is trained based on historical static optimization factors and historical dynamic optimization factors of known performance problems.

[0112] S306: Confirm a performance optimization scheme of the application program based on the performance score vector.

[0113] In one or more embodiments, S302 includes: calculating an optimization coefficient of the optimization strategy based on the optimization effect of the optimization strategy; wherein the optimization effect is a ratio of performance before optimization to performance after optimization; sorting the optimization coefficients, and selecting the largest n target optimization coefficients; and obtaining the static optimization factor of the application program based on the n target optimization coefficients.

[0114] In one or more embodiments, calculating the optimization coefficient of the optimization strategy based on the optimization effect of the optimization strategy includes: obtaining initial performance data of the application program, wherein the initial performance data represents the performance of the application program before the optimization strategy is implemented; obtaining optimized performance data of the application program, wherein the optimized performance data represents the performance of the application program after the optimization strategy is implemented; and calculating a ratio of the mean values of the initial performance data and the optimized performance data to obtain the optimization coefficient of the optimization strategy.

[0115] Specifically, different optimization strategies will produce different optimization effects. In order to facilitate comparison, different optimization coefficients are set for different optimization strategies in this embodiment, and the optimization coefficient is obtained from the ratio of the effect before optimization to the effect after optimization. Specifically, the calculation formula of the optimization coefficient is as follows:

[0116] wherein, represents an optimization coefficient of the optimization strategy , is a set of performance before optimization, is a mean value of performance after optimization. The optimization coefficient calculated above is sorted, and the embodiment adopts a common Top-n sorting filter to extract the top n optimization coefficients to obtain a static optimization factor .

[0117] In one or more embodiments, S304 comprises: comparing the system indicator information with the preset indicators to obtain the deviation degree values of each system indicator information; comparing the behavior sequence with the pre-established performance problem feature library to confirm that the coincidence degree of the behavior sequence and the problem behaviors in the performance problem feature library reaches a preset threshold; and calculating the dynamic optimization factor of the application program based on the deviation degree values of each system indicator information.

[0118] Specifically, the system indicator information is quantified into specific numerical values or states, such as CPU usage represented by percentage, memory consumption represented by MB, data packet sending represented by data size, and the like. The quantified system indicator information is input into a dynamic feature model, and the dynamic feature model compares and analyzes the real-time system indicator information of the application program with the preset indicators to check whether the current system running state is consistent with the expectation, i.e., whether there is deviation or abnormality, and finally obtains the deviation degree values of each system indicator information. The performance problem feature library records a large number of verified "problem behavior sequences" that have a significant impact on performance, and therefore comparing the behavior sequence with the pre-established performance problem feature library can quickly identify whether the current behavior sequence is similar to (such as the coincidence degree reaching a preset threshold) some problem behaviors in the performance problem feature library to determine whether the application program has problems during running and which aspect of the problem.

[0119] In one or more embodiments, comparing the system indicator information with the preset indicators to obtain the deviation degree values of each system indicator information comprises: quantifying the system indicator information into system indicator data; inputting the system indicator data into a pre-established dynamic feature model, wherein the dynamic feature model includes historical system indicator information, and the dynamic feature model generates the preset indicators according to the historical system indicator information; and obtaining the deviation degree values of each system indicator information through the dynamic feature model.

[0120] In one or more embodiments, the establishment process of the performance problem feature library comprises: obtaining the historical behavior sequence of the application program and the historical system indicator information corresponding to the historical behavior sequence; The key behaviors in the historical behavior sequence are taken as sequence points, the historical behavior sequence is recursively classified to obtain a plurality of sub-sequences, wherein the key behaviors are behaviors that have a deep impact on the performance of the application, and each sub-sequence includes one sequence; The plurality of sub-sequences are divided into impact sequences and normal sequences; A target sequence that affects the performance of the application is obtained from the normal sequences, and the target sequence is added to the impact sequences; Based on m target impact sequences with the largest impact on the performance of the application in the impact sequences, a performance problem feature library is generated.

[0121] In one or more embodiments, S305 includes: weighting and fusing the static optimization factor and the dynamic optimization factor to obtain a comprehensive feature factor; and inputting the comprehensive feature factor into a pre-trained deep learning model to obtain a performance score vector output by the deep learning model, wherein the performance score vector is used to represent the similarity between the performance state of the application and the known performance problems.

[0122] Specifically, the calculation formula of the comprehensive feature factor is:

[0123] wherein, represents the comprehensive feature factor, represents the static optimization factor, represents the dynamic optimization factor, represents a weight coefficient.

[0124] In one or more embodiments, S306 includes: confirming the severity of the current performance problem affecting the application based on the performance score vector; and confirming the performance optimization scheme of the application according to the severity.

[0125] In one or more embodiments, confirming the severity of the current performance problem affecting the application based on the performance score vector includes: when the performance score vector reaches a first threshold, confirming that the severity of the current performance problem affecting the application is mild; when the performance score vector reaches a second threshold, confirming that the severity of the current performance problem affecting the application is moderate; and when the performance score vector reaches a third threshold, confirming that the severity of the current performance problem affecting the application is severe; wherein the first threshold, the second threshold, and the third threshold increase.

[0126] For example, the first threshold is 0.7, the second threshold is 0.8, and the third threshold is 0.9. When the similarity is greater than or equal to 0.7, the severity of the current performance problem affecting the application is mild (there may be some impact). When the similarity is greater than or equal to 0.8, the severity is confirmed to be moderate (there is an impact). When the similarity is close to 0.9, the severity is confirmed to be severe (serious impact).

[0127] In one or more embodiments, further comprising: constructing an optimization strategy matrix based on the optimization strategies; wherein the optimization strategy matrix comprises elements determined based on a number of the optimization strategies, and the optimization strategy matrix is used to represent the optimization strategies contained by the application program.

[0128] The optimization strategy matrix is as follows:

[0129] The optimization strategy matrix is represented as Gn x k, the number of strategies is n, representing the application program whether the optimization strategy is contained If yes, the value is 1, otherwise the value is 0. Through the optimization strategy matrix, it can be directly known which optimization strategies are contained by the application program.

[0130] According to another aspect of the embodiments of the present disclosure, an application program performance optimization device is provided, as shown in Figure 4 The device comprises: An analysis module 401 configured to obtain source code of an application program, perform exception analysis on the source code, and determine a plurality of optimization strategies; A static optimization factor generation module 402 configured to obtain static optimization factors of the application program based on optimization effects of the optimization strategies; A running monitoring module 403 configured to obtain running logs of the application program when the application program is running, analyze the running logs, and obtain system index information corresponding to a behavior sequence, wherein the behavior sequence comprises a plurality of behaviors affecting performance of the application program, and the index information represents a performance state of the application program when the behavior sequence occurs; A dynamic optimization factor generation module 404 configured to compare the system index information with preset indexes, and obtain dynamic optimization factors of the application program; A similarity calculation module 405 configured to analyze the static optimization factors and the dynamic optimization factors by using a pre-trained deep learning model, and confirm a performance score vector of the application program; wherein the deep learning model is trained based on historical static optimization factors and historical dynamic optimization factors of known performance problems; An optimization module 406 configured to confirm a performance optimization scheme of the application program based on the performance score vector.

[0131] In one or more embodiments, the static optimization factor generation module 402 is configured to: calculate optimization coefficients of the optimization strategies based on optimization effects of the optimization strategies; wherein the optimization effect is a ratio of performance before optimization to performance after optimization; sort the optimization coefficients, and select the largest n target optimization coefficients; The static optimization factor of the application program is obtained based on the n target optimization coefficients.

[0132] In one or more embodiments, the dynamic optimization factor generation module 404 is configured to: compare the system index information with the preset index to obtain a deviation degree value of each system index information; compare the behavior sequence with the pre-established performance problem feature library to confirm that the coincidence degree of the behavior sequence and the problem behavior in the performance problem feature library reaches a preset threshold; calculate the dynamic optimization factor of the application program based on the deviation degree value of each system index information.

[0133] In one or more embodiments, the similarity calculation module 405 is configured to: weight and fuse the static optimization factor and the dynamic optimization factor to obtain a comprehensive feature factor; input the comprehensive feature factor into a pre-trained deep learning model to obtain a performance score vector output by the deep learning model, wherein the performance score vector is used to represent the similarity between the performance state of the application program and the known performance problem.

[0134] In one or more embodiments, the optimization module 406 is configured to: confirm the severity of the influence of the current performance problem on the application program based on the performance score vector; and confirm the performance optimization scheme of the application program according to the severity.

[0135] The application program performance optimization is also configured to: after obtaining the source code of the application program, perform exception analysis on the source code, determine a plurality of optimization strategies, and construct an optimization strategy matrix based on the optimization strategies; wherein the number of elements included in the optimization strategy matrix is determined based on the number of optimization strategies, and the optimization strategy matrix is used to represent the optimization strategies included in the application program.

[0136] In one or more embodiments, the dynamic optimization factor generation module 404 is further configured to: quantify the system index information into system index data; input the system index data into a pre-established dynamic feature model, wherein the dynamic feature model includes historical system index information, and the dynamic feature model generates the preset index according to the historical system index information; obtain the deviation degree value of each system index information through the dynamic feature model.

[0137] In one or more embodiments, the optimization module 406 is further configured to: when the performance score vector reaches a first threshold, confirm that the severity of the influence of the current performance problem on the application program is mild; When the performance score vector reaches the second threshold value, it is confirmed that the current performance problem has a moderate degree of influence on the application program; When the performance score vector reaches the third threshold value, it is confirmed that the current performance problem has a severe degree of influence on the application program; wherein the first threshold value, the second threshold value and the third threshold value are increasing.

[0138] In one or more embodiments, the static optimization factor generation module 402 is further configured to: obtain initial performance data of the application program, wherein the initial performance data represents performance of the application program before the optimization strategy is implemented; obtain post-optimization performance data of the application program, wherein the post-optimization performance data represents performance of the application program after the optimization strategy is implemented; calculate a ratio of the mean of the initial performance data and the post-optimization performance data to obtain an optimization coefficient of the optimization strategy.

[0139] The application program performance optimization apparatus provided by the embodiments of the present disclosure and the application program performance optimization method provided by the embodiments of the present disclosure have the same beneficial effects as the method they employ, run or implement.

[0140] The embodiments of the present disclosure further provide a computer device for executing the above-mentioned application program performance optimization method. Please refer to Figure 5 which shows a schematic diagram of a computer device provided by some embodiments of the present disclosure. As shown in Figure 5 The computer device 5 includes a processor 500, a memory 501, a bus 502 and a communication interface 503, wherein the processor 500, the communication interface 503 and the memory 501 are connected through the bus 502; the memory 501 stores a computer program which can run on the processor 500, and the processor 500 runs the computer program to execute the application program performance optimization method provided by any one of the preceding embodiments of the present disclosure.

[0141] The memory 501 can include a high-speed random access memory (RAM) and can also include a non-volatile memory such as at least one disk memory. The communication between the apparatus network element and at least one other network element is realized through at least one communication interface 503 (which can be wired or wireless), and the internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0142] The bus 502 can be an ISA bus, a PCI bus, an EISA bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, and the like. The memory 501 is configured to store programs, and the processor 500 executes the programs after receiving execution instructions. The application program performance optimization method disclosed in any of the embodiments of the present disclosure can be applied to the processor 500 or implemented by the processor 500.

[0143] The processor 500 can be an integrated circuit chip with processing capability. In the implementation process, each step of the above method can be completed by an integrated logic circuit or an instruction in the form of software in the processor 500. The processor 500 described above can be a general processor, including a central processing unit (CPU), a network processor (NP), and the like; or can be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Each method, step, and logic block disclosed in the embodiments of the present disclosure can be implemented or executed. The general processor can be a microprocessor or any conventional processor. The steps of the method disclosed in combination with the embodiments of the present disclosure can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, or other mature storage media in the art. The storage medium is located in the memory 501, and the processor 500 reads the information in the memory 501 and combines the hardware to complete the steps of the above method.

[0144] The computer device provided by the embodiments of the present disclosure and the application program performance optimization method provided by the embodiments of the present disclosure have the same beneficial effects as the method adopted, run, or implemented by them.

[0145] The embodiments of the present disclosure also provide a computer readable storage medium corresponding to the application program performance optimization method provided by the preceding embodiments. The computer readable storage medium is an optical disc, and a computer program (i.e., a computer program product) is stored on the optical disc. When the computer program is run by a processor, the application program performance optimization method provided by any of the preceding embodiments is executed.

[0146] It should be noted that examples of the computer-readable storage medium can also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other optical, magnetic storage media, and the like, which will not be listed one by one here.

[0147] The computer-readable storage medium provided by the above embodiments of the present disclosure has the same beneficial effects as the method for optimizing application performance provided by the embodiments of the present disclosure, and has the same beneficial effects as the method adopted, run or implemented by the application stored therein.

[0148] The embodiments of the present disclosure also provide a computer program product, please refer to Figure 6 The computer program product 600 carries a program source code, that is, a computer program 601, and the instructions included in the computer program 601 can be used to execute the steps of the application performance optimization method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0149] The computer program product can be specifically implemented by hardware, software or a combination thereof. In an optional embodiment, the computer program product is specifically embodied as a computer storage medium, and in another optional embodiment, the computer program product is specifically embodied as a software product, such as a software development kit (Software Development Kit, SDK) and the like.

[0150] The basic principles of the present disclosure are described above in combination with specific embodiments, but it should be noted that the advantages, advantages, effects and the like mentioned in the present disclosure are only examples and not limitations, and these advantages, advantages, effects and the like cannot be considered as the must-have of each embodiment of the present disclosure. In addition, the above specific details are only for the purpose of example and understanding, and are not limited to the specific details disclosed above for the implementation of the present disclosure.

[0151] The block diagrams of devices, apparatuses, equipment, systems referred to in the present disclosure are merely illustrative examples and are not intended to require or imply that the connection, arrangement, configuration must be as shown in the block diagrams. These devices, apparatuses, equipment, systems can be connected, arranged, configured in any manner as will be appreciated by those skilled in the art. Words such as "include," "contain," "have," etc. are open-ended words that are to be interpreted to mean "including but not limited to," and are to be interpreted not to exclude other items. The words "or" and "and" as used herein are to be interpreted as the word "and / or," and are to be interpreted not to exclude other items. The word "such as" as used herein is to be interpreted as the phrase "such as but not limited to," and is to be interpreted not to exclude other items.

[0152] In addition, as used herein, the "or" as used in the context "at least one of A, B, or C" : means A or B or C or any combination thereof. Further, the phrase "example of" is not meant to be limiting in terms of the examples described. For example, the phrase "example of A, B, or C" means A or B or C, or any combination thereof.

[0153] It is also important to note that the systems and methods of the present disclosure can be embodied in a variety of forms including, but not limited to, a data processor, a computer program product, a computer, one or more tangible computer readable storage devices, one or more computer-implemented methods, information, or a bit of information. Additionally the systems and methods of the present disclosure can be embodied as one or more computers or computer implementations that include one or more processors or one or more memory modules.

[0154] Various changes, modifications and alterations in the teachings and techniques described herein can be made without departing from the teachings that are defined by the appended claims. Further, the scope of the claims of the present disclosure is not limited to the specific aspects described herein. Processes, machines, manufacture, compositions of matter, means, methods, or steps, presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Accordingly, the appended claims include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.

[0155] The above description of the disclosed aspects is meant to be illustrative of the application and not limiting. Various modifications of the aspects will be apparent to those with ordinary skill in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0156] The foregoing description has been presented for the purposes of illustration and description. Furthermore, the description is not intended to limit the embodiments of the disclosure to the forms disclosed herein. Although the various example aspects and embodiments have been described herein with regard to particular aspects and embodiments, those skilled in the art will recognize that certain modifications, changes, substitutions, additions and sub-combinations can be made without departing from the spirit of the disclosure.

Claims

1. An application performance optimization method, characterized by, The method comprises the following steps: Obtaining the source code of an application, performing exception analysis on the source code, and determining a plurality of optimization strategies; Based on the optimization effect of the optimization strategy, obtaining the static optimization factor of the application; During the running of the application, obtaining the running log of the application, analyzing the running log, obtaining the system index information corresponding to the behavior sequence, wherein the behavior sequence comprises a plurality of behaviors affecting the performance of the application, and the index information represents the performance state of the application when the behavior sequence occurs; Comparing the system index information with the preset index to obtain the dynamic optimization factor of the application; Using a pre-trained deep learning model to analyze the static optimization factor and the dynamic optimization factor to confirm the performance score vector of the application; wherein the deep learning model is trained based on the historical static optimization factor and the historical dynamic optimization factor of the known performance problem; Based on the performance score vector, confirming the performance optimization scheme of the application.

2. The application performance optimization method of claim 1, wherein, Based on the optimization effect of the optimization strategy, obtaining the static optimization factor of the application, comprising: Based on the optimization effect of the optimization strategy, calculating the optimization coefficient of the optimization strategy; Sorting the optimization coefficients and selecting the largest n target optimization coefficients; Based on the n target optimization coefficients, obtaining the static optimization factor of the application.

3. The application performance optimization method of claim 1, wherein, Comparing the system index information with the preset index to obtain the dynamic optimization factor of the application, comprising: Comparing the system index information with the preset index to obtain the deviation degree value of each system index information; Comparing the behavior sequence with the pre-established performance problem feature library to confirm that the coincidence degree of the behavior sequence and the problem behavior in the performance problem feature library reaches a preset threshold; Based on the deviation degree value of each system index information, calculating the dynamic optimization factor of the application.

4. The application performance optimization method of claim 3, wherein, The establishment process of the performance problem feature library comprises: Obtaining the historical behavior sequence of the application and the historical system index information corresponding to the historical behavior sequence; Taking the key behavior in the historical behavior sequence as a sequence point, recursively classifying the historical behavior sequence to obtain a plurality of sub-sequences, wherein the key behavior is a behavior with a deep influence on the performance of the application, and each sub-sequence comprises a sequence; Dividing the plurality of sub-sequences into an impact sequence and a normal sequence; Obtaining a target sequence in the normal sequence that affects the performance of the application, and adding the target sequence to the impact sequence; Based on the m target impact sequences in the impact sequence with the largest influence on the performance of the application, generating a performance problem feature library.

5. The application performance optimization method of claim 1, wherein, Using a pre-trained deep learning model to analyze the static optimization factor and the dynamic optimization factor to confirm the performance score vector of the application, comprising: Weighted fusion of the static optimization factor and the dynamic optimization factor to obtain a comprehensive feature factor; input the comprehensive feature factor into a pre-trained deep learning model to obtain a performance score vector output by the deep learning model, wherein the performance score vector is used to represent a similarity between a performance state of the application program and a known performance problem.

6. The application performance optimization method of claim 1, wherein, Based on the performance score vector, the performance optimization scheme of the application program is confirmed, including: Based on the performance score vector, the severity of the current performance problem affecting the application program is confirmed. According to the severity, the performance optimization scheme of the application program is confirmed.

7. The application performance optimization method of claim 1, wherein, After obtaining the source code of the application program, performing abnormal analysis on the source code, and determining a plurality of optimization strategies, the method further includes: Based on the optimization strategy, an optimization strategy matrix is constructed. Wherein, the number of elements included in the optimization strategy matrix is determined based on the number of optimization strategies, and the optimization strategy matrix is used to represent the optimization strategies contained in the application program.

8. The application performance optimization method of claim 3, wherein, The system indicator information is compared with the preset indicator to obtain the deviation degree value of each system indicator information, including: Quantify the system indicator information into system indicator data; The system indicator data is input into a pre-established dynamic feature model, wherein the dynamic feature model includes historical system indicator information, and the dynamic feature model generates a preset indicator according to the historical system indicator information; The deviation degree value of each system indicator information is obtained through the dynamic feature model.

9. The application performance optimization method of claim 6, wherein, Based on the performance score vector, the severity of the current performance problem affecting the application program is confirmed, including: When the performance score vector reaches a first threshold, it is confirmed that the severity of the current performance problem affecting the application program is mild; When the performance score vector reaches a second threshold, it is confirmed that the severity of the current performance problem affecting the application program is moderate; When the performance score vector reaches a third threshold, it is confirmed that the severity of the current performance problem affecting the application program is severe; Wherein, the first threshold, the second threshold and the third threshold are increasing.

10. The application performance optimization method of claim 2, wherein, Based on the optimization effect of the optimization strategy, the optimization coefficient of the optimization strategy is calculated, including: Obtain the initial performance data of the application program, wherein the initial performance data represents the performance of the application program before implementing the optimization strategy; Obtain the optimized performance data of the application program, wherein the optimized performance data represents the performance of the application program after implementing the optimization strategy; Calculate the ratio of the mean values of the initial performance data and the optimized performance data to obtain the optimization coefficient of the optimization strategy.

11. An application performance optimization apparatus, characterized by comprising: Including: The analysis module is used to obtain the source code of the application program, perform abnormal analysis on the source code, and determine a plurality of optimization strategies. The static optimization factor generation module is used to obtain the static optimization factor of the application program based on the optimization effect of the optimization strategy. A running monitoring module is configured to acquire a running log of the application program when the application program is running, analyze the running log, and acquire system index information corresponding to a behavior sequence, wherein the behavior sequence includes a plurality of behaviors affecting the performance of the application program, and the index information represents the performance state of the application program when the behavior sequence occurs. A dynamic optimization factor generation module is configured to compare the system index information with preset indexes, and acquire a dynamic optimization factor of the application program. A similarity calculation module is configured to analyze the static optimization factor and the dynamic optimization factor by using a pre-trained deep learning model, and confirm a performance score vector of the application program, wherein the deep learning model is trained based on historical static optimization factors and historical dynamic optimization factors of known performance problems. An optimization module is configured to confirm a performance optimization scheme of the application program based on the performance score vector.

12. A computer-embedded device comprising a memory, a processor and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the method of any one of claims 1-10.

13. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-10.

14. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-10.

Images