CN108319974B

CN108319974B - Data processing method, data processing device, storage medium and electronic device

Info

Publication number: CN108319974B
Application number: CN201810061270.1A
Authority: CN
Inventors: 刘江冬; 洪楷; 徐士立; 李孝宁; 吴海洋; 刘专; 胡凡
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2018-01-22
Filing date: 2018-01-22
Publication date: 2022-10-28
Anticipated expiration: 2038-01-22
Also published as: CN108319974A

Abstract

The invention discloses a data processing method, a data processing device, a storage medium and an electronic device. Wherein, the method comprises the following steps: determining the target probability of the client in the stuck state in the process of running in the first time period according to the target model and the current state data of the client; and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state of the client in the process of running in a first time period is lower than the target probability. The invention solves the technical problem of low operation efficiency of the client in the related technology.

Description

Data processing method, data processing apparatus, storage medium, and electronic apparatus

Technical Field

The present invention relates to the field of computers, and in particular, to a data processing method, apparatus, storage medium, and electronic apparatus.

Background

At present, the client end is blocked in the operation process, for example, the phenomena of picture stagnation, discontinuous sound and the like occur, the client end cannot normally operate, and the user experience is influenced.

When dealing with the client stuck problem, it is common to optimize the individual clients at the code level, maintain the parameters in software while the system is running, or modify the parameters directly in the hard disk firmware. However, the above method is complex to operate, cannot modify fixed parameters, is only processed for a specific client, and is not suitable for multiple clients, so that effective measures cannot be taken to avoid the problem that the client is stuck and the operating efficiency of the client is low.

For the problem of low operation efficiency of the client, no effective solution has been proposed at present.

Disclosure of Invention

Embodiments of the present invention provide a data processing method, an apparatus, a storage medium, and an electronic apparatus, so as to at least solve the technical problem of low operating efficiency of a client in the related art.

According to an aspect of an embodiment of the present invention, there is provided a data processing method. The data processing method comprises the following steps: training the initial training model by using the target state data to obtain a trained target model; determining the target probability of a stuck state occurring in the process of the client operating in a first time period according to a target model and current state data of the client, wherein the current state data comprise state data generated by the client and/or a terminal for installing the client in the process of the client operating in a current second time period, and the second time period is earlier than the first time period; and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state of the client in the process of running in a first time period is lower than the target probability.

According to another aspect of the embodiment of the invention, a data processing device is also provided. The data processing apparatus includes: the determining unit is used for determining the target probability of the stuck state occurring in the process of the client operating in the first time period according to the target model and the current state data of the client, wherein the current state data comprises state data generated by the client and/or a terminal for installing the client in the process of the client operating in the current second time period, and the second time period is earlier than the first time period; and the processing unit is used for processing the current state data of the client and/or performing target operation on at least one of the client and the terminal under the condition that the target probability is greater than or equal to a first target threshold value, so that the probability of the occurrence of a stuck state in the process of the client operating in a first time period is lower than the target probability.

According to another aspect of the embodiments of the present invention, there is also provided a storage medium. The storage medium has stored therein a computer program, wherein the computer program is arranged to perform the data processing method of an embodiment of the invention when executed.

According to another aspect of the embodiment of the invention, an electronic device is also provided. The electronic device comprises a memory in which a computer program is stored and a processor arranged to execute the data processing method of an embodiment of the invention by means of the computer program.

In the embodiment of the invention, the target probability of the stagnation state occurring in the process of the client operating in the first time period is determined through the target model and the current state data of the client; and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state in the process of the client operating in the first time period is lower than the target probability. Due to the fact that the target probability of the stuck state occurring in the process of the client running in the first time period is determined through the target model and the current state data of the client, the purpose of reducing the probability of the stuck state occurring in the client by taking effective measures is achieved, the problem that the stuck state of the client cannot be avoided by taking effective measures is avoided, the technical effect of improving the running efficiency of the client is achieved, and the technical problem that the running efficiency of the client in the related technology is low is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

FIG. 1 is a schematic diagram of a hardware environment for a data processing method according to an embodiment of the present invention;

FIG. 2 is a flow chart of a method of data processing according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a data process according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of a player's overall katon odds according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a single player stuck distribution according to an embodiment of the present invention;

FIG. 6 is a diagram illustrating the training results of a decision tree model according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of an interface for optimizing a game client according to an embodiment of the present invention;

FIG. 8 is a schematic diagram of a data processing apparatus according to an embodiment of the present invention; and

fig. 9 is a block diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Moreover, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

According to an aspect of an embodiment of the present invention, an embodiment of a data processing method is provided.

Alternatively, in the present embodiment, the data processing method described above may be applied to a hardware environment formed by the server 102 and the terminal 104 as shown in fig. 1. Fig. 1 is a schematic diagram of a hardware environment of a data processing method according to an embodiment of the present invention. As shown in fig. 1, a server 102 is connected to a terminal 104 via a network including, but not limited to: the terminal 104 is not limited to a PC, a mobile phone, a tablet computer, etc. in a wide area network, a metropolitan area network, or a local area network. The data processing method according to the embodiment of the present invention may be executed by the server 102, the terminal 104, or both the server 102 and the terminal 104. The data processing method of the embodiment of the present invention executed by the terminal 104 may also be executed by a client installed thereon.

Fig. 2 is a flow chart of a data processing method according to an embodiment of the present invention. As shown in fig. 2, the method may include the steps of:

step S202, determining the target probability of the client in the stuck state in the process of running in the first time period according to the target model and the current state data of the client.

In the technical solution provided in the above step S202 of the present application, a target probability of a stuck state occurring during a process of the client operating in a first time period is determined through a target model and current state data of the client, where the current state data includes state data generated by the client and/or a terminal on which the client is installed during a process of the client operating in a current second time period, and the second time period is earlier than the first time period.

In this embodiment, a client may have a stuck state during the running process, for example, the client has a picture stagnation, a sound discontinuity, and the like. Optionally, the client is a game client, corresponding to the game server, and is used to connect to the server to provide a program of local service for the game player. In the running structure of the game client, the game client may be stuck, for example, in the running process of the game, rendering of the game picture is delayed or stopped, so that the game picture is discontinuous or blurred, and the game stuck may seriously affect the game experience of the player.

The client of the embodiment may be installed on a terminal, and the terminal may be a terminal device such as a smart phone, a tablet computer, a palm computer, and a mobile internet device, which is not limited herein.

The target model of this embodiment may be a decision tree model trained using user history data for determining in real time the probability that a player is currently likely to have a game stuck. The method comprises the steps of obtaining current state data of a client, and determining target probability of a stuck state occurring in the process of the client operating in a first time period through a target model and the current state data of the client, wherein the first time period, namely a prediction time period, is a time period in which the stuck state occurring in the process of the client operating needs to be predicted, and can be a time period after the current time. The target probability is used to indicate a tendency of the client to exhibit a stuck state during the first time period of operation, i.e., a likelihood of exhibiting a stuck state.

The current status data of this embodiment includes status data generated by the client and/or the terminal during the process that the client operates in the current second time period, where the second time period is a time period in which the current status data of the client has been generated currently, and is earlier than the first time period, and may be a time period consecutive to the second time period. Optionally, in the process that the client is running in the current Second time period, the usage amount of a Random Access Memory (RAM) of the terminal, the usage rate of a Central Processing Unit (CPU) of the terminal, the battery temperature of the terminal, a scene Identifier (ID) of the client, and a Frame rate (FPS) of the client in running are obtained. The current state data may be real-time game state data.

And step S204, under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state of the client in the process of running in a first time period is lower than the target probability.

In the technical solution provided in the foregoing step S204 of the present application, after determining a target probability that a stuck state occurs during a process in which the client operates in a first time period, under a condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that a probability that the stuck state occurs during the process in which the client operates in the first time period is lower than the target probability.

After the target probability of the client in the stuck state in the process of running in the first time period is determined, whether the target probability is larger than or equal to a first target threshold value is judged. The first target threshold is a critical probability value for taking optimization measures for the client. If the target probability is judged to be greater than or equal to the first target threshold, it is indicated that the probability of the stuck state occurring in the process of the client operating on the first time period is higher, in order to avoid the stuck state occurring in the process of the client operating on the third time end, a certain optimization measure may be taken, and the current state data of the client is processed, so that the probability of the stuck state occurring in the process of the client operating on the first time period is lower than the target probability, for example, when the current state data of the client is processed, the image quality of the client may be reduced, a prompt message may be displayed to guide a user operation to reduce the image quality of the client, so as to improve the fluency of the image of the client, the purpose of reducing the occurrence probability of the client may be achieved by reducing the resolution of the image of the client, and a target operation may be performed on the client and/or the terminal, for example, a target operation such as suggesting to close the client, suggest to directly close the client, suggest to close the terminal, suggest to directly restart the client, suggest to clear the memory of the terminal, and the like, which is only by way of example, and without any limitation.

Optionally, in the embodiment, a certain operation authority is invoked by a co-operating manufacturer to process a large core and a small core of a terminal for installing the client, so as to reduce the probability of the client jamming; the method can also forcibly preempt and operate the Application (APP for short) through a background operation strategy to reduce the probability of the client end jamming, and can improve the operation efficiency of the mobile game client end when the client end is the mobile game client end.

Through the steps S202 to S204, the target probability of the stagnation state occurring in the running process of the client in the first time period is determined through the target model and the current state data of the client; and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state in the process of the client operating in the first time period is lower than the target probability. The target probability of the stuck state occurring in the process of the client operating in the first time period is determined through the target model and the current state data of the client, so that the aim of reducing the probability of the stuck state occurring in the client by taking effective measures is fulfilled, the problem that the stuck state of the client cannot be avoided by taking effective measures is avoided, the technical effect of improving the operating efficiency of the client is achieved, and the technical problem of low operating efficiency of the client in the related technology is solved.

As an optional implementation manner, before determining, by using the target model and the current state data of the client, a target probability that the client has a stuck state during the operation over the first time period, the method further includes: acquiring target state data, wherein the target state data comprises state data generated by the client and/or a terminal for installing the client under the condition that a stuck state occurs in the process that the client operates in a third time period which is earlier than the second time period; and training the initial training model by using the target state data to obtain a trained target model.

Before determining the target probability of the client in the stuck state in the process of running in the first time period through the target model and the current state data of the client, the target state data can be obtained, so that the initial training model is trained by using the target state data, and the trained target model is obtained. The third period of time, i.e., the history period of time, of this embodiment may be a period of time before the current state data is generated, earlier than the second period of time, and may be continuous with the second period of time. In the case where the client is in the stuck state during the operation over the past third time period, the state data generated by the client and/or the terminal is the target state data, that is, the target state data is the historical operation state data of the client in the stuck state, and may be historical game state data. Optionally, the target state data of this embodiment includes RAM usage of the terminal, CPU usage of the terminal, battery temperature of the terminal, scene identification ID of the client, frame rate FPS of the client at runtime, and the like. The scene ID of the client can be the game scene ID of the client, the FPS of the client during operation is used for measuring the measurement of the display frame number, the measurement unit of the FPS is 'display frame number per second' or 'Hertz', the FPS can be used for describing how many frames are played by video, electronic drawing or games per second, and the FPS can be the game FPS of the client during operation.

Optionally, in the embodiment, from the historical state data generated by the client and/or the terminal, the historical state data of the client in the stuck-in state is marked by a stuck-in calculation formula, and the historical state data of the marked client in the stuck-in state may be the target state data, so that quantification of the stuck-in state of the client is achieved.

Optionally, in this embodiment, the initial training model is an initial training model, and may be an initial neural network model, and the initial neural network model is described based on a mathematical model of a neuron, and the neuron is established by the state data of the client, which is collected at the beginning. After the target state data is acquired, the embodiment takes the target state data as a target sample for training an initial training model. In the actual data processing process, after acquiring the original data, feature selection is required to be performed first, and then the model is trained. And training the initial training model through the state data with the relatively high relation with the stuck state of the client to obtain the trained target model.

Optionally, the embodiment may also periodically update the trained target model, so that the target model is suitable for predicting the probability of occurrence of seizure according to the real-time state data of the client.

Alternatively, the target model in this embodiment may be a supervised classification model, and may be a decision tree model, where the decision tree model is a prediction model, and is also a tree-like decision diagram with probability results, and is used to represent a mapping between object attributes and object values, each node in the decision tree model is used to represent a judgment condition of object attributes, and its branch represents an object meeting the node condition, and a leaf node of the tree represents a prediction result to which the object belongs.

As an optional implementation, the acquiring the target state data includes: acquiring historical state data, wherein the historical state data comprises state data generated by the client and/or a terminal for installing the client in the process that the client runs on a third time period which is past; in the historical state data, target state data at a target frame rate is marked, wherein the target frame rate is the frame rate of the client when the client is in the stuck state.

In this embodiment, historical state data is obtained, which may be historical game state data including state data generated by the client and/or the terminal in which the client is installed during the client's operation over the elapsed third time period, for example, including RAM usage of the terminal, CPU usage of the terminal, battery temperature of the terminal, scene identification ID of the client, FPS of the client during operation, and the like.

In general, when the client is in the stuck state, it can only be known that the client has stuck within a certain period of time, and the severity of the stuck occurrence cannot be accurately quantified. After the historical state data is obtained, the target state data at the target frame rate is marked, the target state data can be marked through a stuck-at formula based on an FPS value of the client, that is, a data point of the client in the stuck-at state is marked, wherein the target frame rate is the frame rate of the client in the stuck-at state, the initial training model is trained by using the target state data to obtain a trained target model, the target probability of the client in the stuck-at state in the process of operating in the first time period is determined through the target model and the current state data of the client, and the current state data of the client is processed and/or at least one of the client and the terminal is subjected to target operation when the target probability is greater than or equal to a first target threshold, so that the probability of the client in the stuck-at state in the process of operating in the first time period is lower than the target probability, thereby improving the operating efficiency of the client.

As an optional implementation manner, after the target state data at the target frame rate is marked, the method further includes: selecting first target characteristic data from the target state data, wherein the first target characteristic data is state data, in the historical state data, of which the correlation degree with the stuck state of the client is higher than a second target threshold; training the initial training model by using the target state data, and obtaining the trained target model comprises the following steps: and training the initial training model by using the first target characteristic data in the target state data to obtain a trained target model.

In this embodiment, the characteristic factors causing the client to be in the stuck state are many, for example, the CPU utilization rate is too high, the memory is insufficient, the battery temperature is too high, the game scene is switched, and the mobile phone power is insufficient. However, the above feature factors are not all critical factors that affect the occurrence of the stuck state at the client, and if the initial training model is trained by all the above feature factors, the difficulty and complexity of training are increased, thereby affecting the efficiency of training the initial training model. Therefore, it is necessary to select features related to the occurrence of the stuck state of the client from all the above feature factors, and remove irrelevant features, so that not only the problem of dimension disaster can be effectively avoided, but also the task difficulty of data processing can be reduced.

After the target state data at the target frame rate is marked, preprocessing is performed on the target state data, feature selection can be performed on the target state data, and first target feature data is selected from the target state data, where the first target feature data is state data in which the degree of correlation between the target state data and the stuck state of the client is higher than a second target threshold, that is, state data in which the degree of correlation between the target state data and the stuck state of the client is high is selected from the target state data. The second target threshold of this embodiment is a critical value for selecting, from the target state data, the first target feature data having a higher degree of correlation with the stuck state of the client. Alternatively, the state data in which the degree of correlation with the stuck state of the client in the target state data is lower than the second target threshold may be determined that the correlation with the stuck state of the client is low, and may be removed. After the first target characteristic data is selected from the target state data, the initial training model is trained by using the first target characteristic data in the target state data to obtain a trained target model, namely, the initial training model is trained by using the characteristic data with higher correlation degree with the stuck state of the client to obtain the trained target model, so that the characteristic dimensionality is reduced, the problem of dimensionality disaster is effectively avoided, and the task difficulty of data processing can be reduced.

As an optional implementation, selecting the first target feature data from the target state data includes: selecting candidate feature data from the target state data, wherein the number of the candidate feature data is a first number; evaluating the candidate feature data of the first quantity respectively to obtain evaluation results of the first quantity; obtaining target evaluation results meeting target conditions in the first number of evaluation results; and determining candidate feature data corresponding to the target evaluation result as first target feature data.

In this embodiment, after the target state data is acquired, candidate feature data may be selected from the target state data, where the candidate feature data may be candidate feature subsets, and a first number of the candidate feature data are evaluated respectively, and an information entropy gain of the candidate feature subsets may be acquired to evaluate the candidate feature data, so as to obtain a first number of evaluation results, and further obtain a target evaluation result meeting a target condition in the first number of evaluation results, where the candidate feature data corresponding to the target evaluation result meeting the target condition is an optimal feature subset, and further determine the candidate feature data corresponding to the target evaluation result as the first target feature data.

Optionally, the target state data of this embodiment is a feature set { f } ₁ ，f ₂ ，...，f _n In which f ₁ ，f ₂ ，...，f _n Is the feature data in the feature set, wherein n is a natural number greater than or equal to 1, and is selected from { f ₁ ，f ₂ ，...，f _n The candidate feature data selection method comprises the following steps:

step 1, mixing { f ₁ ，f ₂ ，...，f _n Each feature in f is considered as a candidate feature subset and is paired with f ₁ ，f ₂ ，...，f _n The n candidate feature subsets are evaluated to obtain the evaluation of each candidate feature subsetAnd (6) obtaining the result.

Step 2, obtaining the optimal candidate feature subset { f according to the evaluation result of each candidate feature subset _t And t is more than or equal to 1 and less than or equal to n, and t is a natural number.

Step 3, at { f ₁ ，f ₂ ，...，f _n Divide by { f } _t Selecting a feature { f } from n-1 candidate subsets other than _m Add { f } _t Constitute a candidate subset f containing two features _t ，f _m N-1 candidate subsets can be generated, where 1 ≦ m ≦ n, m ≠ t, and m is a natural number.

Step 4, selecting the optimal candidate subset as { f ] from the generated n-1 candidate subsets _t ，f _s Will then { f } _t ，f _s S is not less than 1 and not more than n, s is not equal to t, and s is a natural number.

Repeating the above steps 3 and 4, for example, at { f } ₁ ，f ₂ ，...，f _n Divide by { f } _s ，f _t In n-2 candidate subsets other than { f }, a feature is selected _m Add { f } _t ，f _s In the method, a candidate subset f containing three features is formed _t ，f _s ，f _m′ N-2 candidate subsets can be generated, where 1 ≦ m '≦ n, m' ≠ t, s. Among the n-2 candidate subsets generated, the optimal candidate subset is selected as f _t ，f _s ，f _s′ Will then { f } _t ，f _s ，f _s′ And the subset is selected as the round, s ' is more than or equal to 1 and less than or equal to n, s ' is not equal to t, s, and s ' is a natural number.

When the selected subset generated by the k +1 round is worse than the selected subset generated by the k round, the generation of the candidate subset is stopped, and the selected subset generated by the k round is used as the optimal feature subset, and the features in the optimal feature subset can be determined as the first target feature data, wherein k is a natural number greater than or equal to 1.

As an alternative implementation, the evaluating the first number of candidate feature data respectively, and obtaining the first number of evaluation results includes: respectively obtaining information entropy gains of a first number of candidate characteristic data to obtain a first number of information entropy gains, wherein the evaluation result comprises the information entropy gains; the acquiring of the target evaluation result meeting the target condition in the first number of evaluation results includes: obtaining information entropy gains larger than a third target threshold value from the first number of information entropy gains; and determining the information entropy gain larger than the third target threshold value as a target evaluation result.

In this embodiment, the information gain entropy of the candidate feature data may be used as a criterion for evaluating the candidate feature data. After the candidate feature data are selected from the target state data, information entropy gains of a first number of the candidate feature data may be obtained, respectively, to obtain the information entropy gains of the first number. For each candidate feature data, its information entropy gain may be calculated based on the training data set as a criterion for evaluating the candidate feature data.

Alternatively, for a given data set D, assume that the proportion of class i samples in the data set D is p _i (i =1,2, \8230 |), wherein | C | is the number of classes. For candidate feature data F, assume that D is divided into V subsets { D) according to its value ¹ ，D ² ，...，D ^V The values of the samples of each subset on the candidate feature data F are the same, so that the information entropy gain of the candidate feature data F can be calculated,

c is the number of classes, e.g., 2,p _i The proportion of the i-th sample is indicated.

In this embodiment, the greater the information entropy gain in the candidate feature data, the more information that is included in the candidate feature data to facilitate classification is meant. After information entropy gains of a first number of candidate characteristic data are respectively obtained, information entropy gains larger than a third target threshold value are obtained in the first number of information entropy gains, the third target threshold value is a critical information entropy gain of the information entropy gains set when whether the information entropy gains are determined to be target evaluation results, the critical information entropy gains are used as basic reference data for determining the target evaluation results, the information entropy gains larger than the third target threshold value are determined to be target evaluation results, namely, the candidate characteristic data corresponding to the information entropy gains larger than the third target threshold value are determined to be first target characteristic data, optionally, the first target characteristic data comprises { game scene ID, mobile phone RAM usage, mobile phone RAM idle amount, mobile phone CPU usage rate and mobile phone battery temperature }, further, target state data at a target frame rate are marked in the first target characteristic data, an initial training model is trained by using the target state data, a target model after training is obtained, and a target card with a probability of occurrence in a first target time period is determined by the target model and current state data of a client; and under the condition that the target probability is greater than or equal to the first target threshold, processing the current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state in the process of the client operating in the first time period is lower than the target probability, and the operating efficiency of the client is improved.

As an optional implementation manner, in the history status data, marking the target status data at the target frame rate includes: acquiring a plurality of frame rates generated by a client in a target time period; acquiring a first weight corresponding to an average frame rate of a plurality of frame rates, wherein the average frame rate is in direct proportion to the first weight, and the first weight is used for indicating the influence proportion of the average frame rate on the hiton state of the client; acquiring a second weight corresponding to the variance of the plurality of frame rates, wherein the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the stuck state of the client; obtaining a target score through the first weight and the second weight, wherein the target score is in direct proportion to the average frame rate of the plurality of frame rates and in inverse proportion to the variance of the plurality of frame rates; and under the condition that the target score is smaller than a fourth target threshold value, determining the frame rate corresponding to the target score as the target frame rate, and marking the target state data at the target frame rate in the historical state data.

In this embodiment, after obtaining the historical state data, the obtaining client generates within the target time periodI.e., acquiring a set of FPSs. Optionally, for a game of a game-on-game type, the multiple frame rates are all FPS values generated in a time sequence within a complete game-on-game, and for a game of a non-game-on-game type, the multiple frame rates are all FPS values generated in a time sequence within the same game scene ID. Optionally, the FPS value set S (FPS acquisition frequency is 5 seconds per point) of the game client is: s = { f ₁ ，f ₂ ，…，f _n }，n＞N _min Wherein, f ₁ ，f ₂ ，…，f _n For representing N FPS values, N > N _min For representing that the number of FPS in the FPS set is more than N only when being collected _min The calculation is performed at the moment, and different settings are provided for different game clients.

After the plurality of frame rates generated by the client in the target time period are obtained, an average frame rate of the plurality of frame rates is obtained, and then a first weight is obtained through the average frame rate, wherein the average frame rate is in direct proportion to the first weight, and the first weight is also the average weight and is used for indicating the influence proportion of the average frame rate on the hiton state of the client.

Alternatively, this embodiment may calculate the average of the FPS throughout the alignment process:

the average of the FPS is used to calculate the average weight:

if mean＜MAAN_THRESHOLD_1：

mean_factor＝MEAN_MIN_FACTOR

else if MEAN_THRESHOLD_1≤mean＜MEAN_THRESHOLD_2：

mean_factor＝MEAN_COEF_1×mean-MEAN_INTERCEPT_1

else：

mean_factor＝MEAN_COEF_2×mean-MEAN_INTERCEPT_2

mean_factor＝min(mean_factor，100)/100

the embodiment may divide the average value of the FPS into three different gears, and the boundary values may be MEAN _ THRESHOLD _1 and MEAN _ THRESHOLD _2, respectively, where the setting of the boundary values needs to consider the highest frame of the game and the rendering characteristics of the game client's own picture. In three gears, the weight calculation parameters may be MEAN _ COEF _ x and MEAN _ interval _ x, where x =1,2.

This embodiment finally ensures that the average weight of the FPS is proportional to the average value of the FPS, and the weight of the average value is dominant in the final stuck score calculation in the lower gear, for example, when MEAN < MEAN _ THRESHOLD _1, the weight of the average value is dominant in the final stuck score calculation, wherein the value of the average weight is in the range of [0.1,1].

After the plurality of frame rates generated by the client in the target time period are acquired, a second weight corresponding to the variance of the plurality of frame rates, namely a variance weight, is acquired, and the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the hiton state of the client.

After the first weight and the second weight are obtained, an object score is obtained through the first weight and the second weight, wherein the object score is proportional to an average frame rate of the plurality of frame rates and inversely proportional to a variance of the plurality of frame rates. The target score amount may be a katon score.

Optionally, after obtaining mean weight mean _ factor, FPS variance sum _ var _ score for the FPS, calton score = max (mean _ factor x (100-sum _ var _ score), 1.0) is calculated.

The katton score calculation formula of this embodiment is required to ensure that the final katton score is proportional to the mean weight of the FPS and inversely proportional to the FPS variance. score has a value range of [1, 100].

After the target SCORE is obtained through the first weight and the second weight, if the target SCORE is smaller than a fourth target THRESHOLD, the frame rate corresponding to the target SCORE is determined as the target frame rate, and the target state data at the target frame rate is marked in the history state data, wherein the fourth target THRESHOLD is a critical SCORE of a frame rate set when determining whether the frame rate is the target frame rate, and is used as basic reference data for determining the frame rate as the target frame rate, for example, a carrton SCORE THRESHOLD value SCORE _ THRESHOLD.

OptionallyAfter the katon score is obtained, the state data of the game client is subjected to a katon flag. When the FPS value corresponding to the state data satisfies SCORE < SCORE _ THRESHOLD, and f _i ＜mean，f _i And when the state data belongs to the S condition, the state data is marked as the Caton state data, so that the marking of the state data is realized. Wherein SCORE _ THRESHOLD is a THRESHOLD for the SCORE of morton, and the specific setting can be determined by combining the actual situation of the game service.

As an optional implementation manner, obtaining the second weight corresponding to the variance of the plurality of frame rates includes: dividing a plurality of frame rates into a plurality of frame rate groups; acquiring the variance of the frame rate in each group of frame rate groups; acquiring a target frame rate group corresponding to the variance larger than a fourth target threshold; and acquiring a second weight corresponding to the variance of the frame rates in the target frame rate group.

In this embodiment, when the second weight corresponding to the variance of the multiple frame rates is obtained, the multiple frame rates are divided into multiple frame rate groups, and optionally, the obtained FPS set S is divided into groups, and S may be divided into N equally _split The amount of the organic compound, for example,

wherein S is _i ＝{f _i1 ，f _i2 ，…f _ik }，

S _i ＝{f _i1 ，f _i2 ，…f _ik }，

N _split The natural number is greater than or equal to 2 and is used for representing the number of groups for equally dividing S, and the number of divided groups can be adjusted according to different game services. Wherein the larger the number of sliced groups, the greater the impact of FPS fluctuations over a short time interval on the final stuck score.

After the variance of the frame rates in each group of frame rate groups is obtained by dividing the plurality of frame rates into a plurality of frame rate groups, the target frame rate group corresponding to the variance larger than a fourth target threshold value is obtained, and the fourth target threshold value is a minimum fluctuation variance which can be set for the game.

Alternatively, this embodiment may measure the FPS fluctuation over a short time by calculating a variance weight:

if variance _i ＞MIN_VAR：

var_score _i ＝min(VAR_COEF×(variance _i -MIN_VAR)，1)

wherein S is _i Is the ith frame rate group comprising multiple frame rates and variance threshold variance _i The variance greater than MIN _ VAR is the minimum fluctuation variance set for the game, i.e. it is considered that in a small time interval, the FPS fluctuation variance greater than MIN _ VAR will have an effect on the morton of the game client, and the variance threshold variance is variable _i The > MIN _ VAR is adjusted according to different services and needs to be combined with actual conditions. The variance weight calculation parameter is VAR _ COEF, which mainly controls the weight of the variance value in the final calton score calculation, and the larger the parameter is, the larger the influence of the variance value on the final calton score is.

After an object frame rate group corresponding to the variance larger than a fourth object threshold value is obtained, a second weight corresponding to the variance of the frame rate in the object frame rate group is obtained, then after an object score is obtained through the first weight and the second weight, under the condition that the object score is smaller than the fourth object threshold value, the frame rate corresponding to the object score is determined as the object frame rate, and object state data under the object frame rate is marked in the historical state data, so that the marking of the state data of the game client is achieved.

As an alternative implementation, in step S202, training an initial training model using the target state data, and obtaining a trained target model includes: determining target state data as object attributes of a target model; determining a result of the client in the stuck state as an object value of a target model, wherein the object attribute and the object value have a mapping relation in the target model; and training the initial training model through the object attributes and the object values to obtain a trained target model, wherein the target state data is critical data for determining that the client is in a stuck state in the trained target model.

The target model of this embodiment is a prediction model, and is also a model with probability added, and may be a decision tree model, and is used for representing a mapping between object attributes and object values, where each node in the target model is used for representing a judgment condition of an object attribute, its branch is used for representing an object meeting the judgment condition of the node, and a leaf node of the target model may be used for representing a prediction result to which the object belongs.

And determining the target state data as the object attribute of the target model, for example, the object attribute of the target model is the characteristics of game scene ID, mobile phone RAM usage, mobile phone RAM idle amount, mobile phone CPU usage rate, mobile phone battery temperature and the like, and determining the result of the client terminal in the stuck state as the object value of the target model, for example, the object value of the target model is whether the client terminal is in the stuck state or not. The purpose of model training is to determine a critical judgment condition of an object attribute when a client terminal is in a stuck state, train an initial training model through the object attribute and an object value to obtain a trained target model, and the target state data of the embodiment is critical data for determining the stuck state of the client terminal in the trained target model.

The target model of this embodiment may be a decision tree model, comprising a root node, a number of internal nodes, and a number of leaf nodes. The leaf nodes correspond to decision results, and each of the other nodes corresponds to an attribute test. The sample set contained by each node is divided into child nodes according to the result of the attribute test. The method comprises the steps that a root node comprises a sample complete set, a judging test sequence corresponds to a path from the root node to each leaf node, a tree with strong generalization capability is generated, the target probability of the blocking state occurring in the process of the client operating in a first time period is determined through a decision tree model and the current state data of the client, the current state data of the client is processed under the condition that the target probability is larger than or equal to a first target threshold value, and/or at least one of the client and a terminal is subjected to target operation, so that the probability of the blocking state occurring in the process of the client operating in the first time period is lower than the target probability, and the operating efficiency of the client is improved.

As an alternative implementation, in step S202, determining, by using the target model and the current state data of the client, a target probability that the client is in a stuck state during the operation over the first time period includes: and determining the target probability of the stuck state of the client in the process of running in the first time period through the target model and the second target state data, wherein the current state data of the client comprises the second target state data, and the second target characteristic data is state data of which the correlation degree with the stuck state of the client in the current state data is higher than a second target threshold value.

Optionally, in this embodiment, the current state data reported by the client includes second target state data, where the second target state data may be state data in which a degree of correlation between the current state data and a stuck state of the client is higher than a second target threshold, and the second target state data is obtained through feature selection, so as to select related state data and remove unrelated state data, and further quickly determine, through a target model, a target probability that the stuck state occurs during the process of the client operating in the first time period, and in a case that the target probability is greater than or equal to the first target threshold, process the current state data of the client and/or perform target operation on at least one of the client and the terminal, so that the probability that the stuck state occurs during the process of the client operating in the first time period is lower than the target probability, and the probability of the client operating is increased.

As an optional implementation manner, after processing the current state data of the client and/or performing the target operation on at least one of the client and the terminal in step S204, the method further includes: acquiring state data of a client in the actual operation process at intervals of target time; and training the target model through the state data of the client in the actual operation process so as to update the target model.

In this embodiment, the target model may be periodically trained, the state data of the client in the actual operation process may be obtained at a target time interval after processing the current state data of the client and/or performing target operation on at least one of the client and the terminal, and then the target model is trained through the state data of the client in the actual operation process to update the target model, so that the target model is suitable for predicting the probability of occurrence of the stuck state according to the real-time state data of the client, and an effective optimization strategy is timely adopted to reduce the occurrence of the stuck state of the client.

The embodiment is a solution for improving the client terminal stuck state, a target model is obtained through training of user historical data, the probability that the client terminal is possibly stuck state is judged in real time through the target model, and if the probability that the client terminal is stuck state is greater than or equal to a set threshold value, an optimization strategy can be adopted, the stuck state of the client terminal is prevented, the operating efficiency of the client terminal is improved, and therefore the user experience is improved.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art will appreciate that the embodiments described in this specification are presently preferred and that no acts or modules are required by the invention.

Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.

The technical solution of the present invention will be described below with reference to preferred embodiments. Specifically, a client is taken as a game client, and a target model is taken as a decision tree model for example.

Fig. 3 is a schematic diagram of a data process according to an embodiment of the present invention. As shown in fig. 3, in the process of running the game client, the data acquisition module acquires game state data, calculates historical game state data of the client by using a katon calculation formula, calculates the historical game state data based on game FPS values, marks game state data belonging to katon in the historical game state data, trains the marked data to obtain a decision tree model, and can perform periodic offline training on the decision tree model to update the decision tree model.

In addition, the game client reports game state data to the data acquisition module in real time, and the data acquisition module receives the game state data reported by the game client, sends the game state data to the trained decision tree model, and simultaneously saves the state data of all the game clients.

And the decision tree model predicts the probability of the client terminal generating the pause according to the received real-time game state data, wherein the probability of the client terminal generating the pause is the pause generation data. And if the probability of the client end jamming is larger than or equal to a certain threshold value, triggering an optimization strategy model, and selecting an optimization strategy for the game client end for optimization. Optionally, when the game client is optimized, the fluency of the game client can be improved by adopting a mode of reducing the image quality of the game client; the method can also reduce the probability of the client end jamming by calling a certain operation authority to process the large and small cores of the terminal for installing the client end through a cooperative manufacturer; the strategy can be operated in the background to forcibly preempt and operate the application, so that the probability of the client terminal being blocked is reduced, and the operating efficiency of the client terminal is improved.

It should be noted that the data acquisition module, the optimization policy module, and the like related to the above-mentioned solution of the embodiment are only core processing modules of the data processing method of the embodiment, and are not all processing modules of the background server.

The scheme of the embodiment can comprise the steps of Katon calculation, feature selection and training of a decision tree model. The following describes the calton calculation.

Generally, when a game client is in a pause, the game client can only know that the game client has paused for a certain period of time, but the severity of the pause cannot be accurately quantified. The decision tree model belongs to a supervised classification model, and before the model is trained, the state data of the game client terminal in the stuck state must be known clearly. In this embodiment, the status data of the player when the player makes a pause is marked by using a pause calculation formula based on FPS, and the calculation process of the pause calculation formula is as follows:

and acquiring an FPS set, wherein the FPS set is used for playing the game of the game pair, generating all FPS values in a complete game pair according to the time sequence for the game of the non-game pair, and generating all FPS values in the same game scene ID according to the time sequence.

Optionally, the FPS value set S (FPS acquisition frequency is 5 seconds per point) of the game client is:

S＝{f ₁ ，f ₂ ，…，f _n }，n＞N _min (1)

wherein f is ₁ ，f ₂ ，…，f _n For representing N FPS values, N > N _min For indicating that the FPS number in the FPS set is larger than N only when the acquisition is carried out _min The calculation is performed only when. With different settings for different game clients.

Calculating the average value of FPS in the whole process of alignment:

the average of the FPS is used to calculate the average weight.

Calculating the average weight mean _ factor of the FPS:

in the embodiment, the average value of the FPS is divided into three different gears, and the boundary values are MEAN _ THRESHOLD _1 and MEAN _ THRESHOLD _2, where the setting of the boundary values needs to consider the rendering characteristics of the highest frame of the game and the picture of the game client itself, where the highest frame of the game may be 30 frames or 60 frames. In three gears, the weight specific calculation parameters are MEAN _ COEF _ x and MEAN _ interval _ x, where x =1,2.

In this embodiment, the acquired FPS sets S are grouped, and S may be equally divided into N _split Parts, namely:

wherein S is _i ＝{f _i1 ，f _i2 ，…f _ik }，

N of this example _split For indicating the number of groups for equally dividing S, for different gamesThe number of service and segmentation groups can be adjusted. Wherein the larger the number of sliced groups, the greater the impact of FPS fluctuations over a short time interval on the final stuck score.

This embodiment measures the fluctuation of the FPS over a short period of time by calculating a variance weight.

Wherein the variance threshold value variance _i The variance threshold variance is considered to be the minimum fluctuation variance set for the game, i.e., it is considered that the FPS fluctuation variance is larger than the MIN _ VAR in a small time interval to have an influence on the mortgage of the game client, and _i the > MIN _ VAR is adjusted according to different services and needs to be combined with actual conditions. The variance weight calculation parameter is VAR _ COEF, which mainly controls the weight of the variance value in the final calton score calculation, and the larger the parameter is, the larger the influence of the variance value on the final calton score is.

After obtaining the mean weight mean _ factor, FPS variance sum _ var _ score for the FPS, calton score is calculated:

score＝max(mean_factor×(100-sum_var_sCore)，1.0) (6)

the katon score calculation formula of this embodiment ensures that the final katon score is proportional to the mean weight of the FPS and inversely proportional to the FPS variance. score has a value range of [1, 100].

After the katon score is obtained, the state data of the game client is subjected to katon tagging. When the FPS value corresponding to the state data meets the condition of the formula (7), the state data is marked as the stuck state data, and therefore the marking of the state data is achieved.

SCORE < SCORE _ THRESHOLD and f _i ＜mean，f _i ∈S (7)

Wherein, SCORE _ THRESHOLD is a THRESHOLD of the katon SCORE, and the specific setting can be determined by combining the actual situation of the game service.

The embodiment adopts the above calculation steps to perform verification on the game service, and the specific verification result is shown in fig. 4 and fig. 5, where fig. 4 is a schematic diagram of the overall katon account ratio of the player according to the embodiment of the invention, and fig. 5 is a schematic diagram of the katon distribution situation of the single player according to the embodiment of the invention. As shown in fig. 4 and 5, the blue indication line is used to indicate the client-side stuck distribution calculated by the frame interval time (millisecond level) when the game client is developed, so that the client-side stuck condition can be accurately measured, and the frame interval time can be acquired at the embedded point in the client-side program due to the high acquisition cost. The green indicating line indicates the katon distribution situation of the client, which is obtained by the katon calculation formula provided in this embodiment. It can be seen that the two lines are in a substantially uniform trend, whether from the entire player or the individual player.

The following describes the selection of features of embodiments of the present invention.

The feature selection of this embodiment is to select a feature having a higher correlation from among a plurality of game state data.

The factors causing the client end jam are many, for example, the CPU utilization rate is too high, the memory is insufficient, the battery temperature is too high, the game scene is switched, the mobile phone power is insufficient, and the like, but the characteristic factors are not all key factors influencing the player jam, and relevant characteristics need to be selected from all the characteristic factors to remove the irrelevant characteristics, so that the problem of dimension disaster can be effectively avoided, and the difficulty of a data processing task can be reduced.

The feature selection is mainly divided into two links: 1) Searching a candidate characteristic subset; 2) And evaluating the candidate feature subset.

1) Candidate feature subset search

For a given set of features f ₁ ，f ₂ ，...，f _n The specific steps of generating the candidate feature subset are as follows:

step a), firstly, regarding each feature as a candidate subset, and evaluating the n candidate single feature subsets, wherein n is a natural number which is more than or equal to 1;

step b) assuming the optimal candidate feature subset generated by step a) as f _t T is more than or equal to 1 and less than or equal to n, and t is a natural number;

step c) selecting a feature { f) from the remaining n-1 candidate subsets _m Add { f } _t Form a candidate subset f containing two features _t ，f _m N-1 candidate subsets can be generated, where 1 ≦ m ≦ n, m ≠ t, and m is a natural number;

assuming that among the n-1 candidate subsets generated in step c), the optimal candidate subset is { f _t ，f _s Will then { f } _t ，f _s S is not less than 1 and not more than n, s is not equal to t, and s is a natural number;

and repeating the steps c) and d), when the selected subset generated by the k +1 round is worse than the selected subset generated by the k round, stopping generating the candidate subset, and taking the selected subset generated by the k round as the optimal feature subset, wherein k is a natural number which is more than or equal to 1.

2) Candidate feature subset evaluation

For a given data set D, assume that the proportion of class i samples in D is p _i (i =1,2, \8230; |) where | C | is the number of classes. For the feature subset F, assume that D is divided into V subsets { D) according to its value ¹ ，D ² ，...，D ^V And F, the samples in each subset have the same value, so that the information entropy gain of the feature subset F can be calculated, as shown in formulas (8) and (9):

wherein, the larger the information Gain (F), the more information that the feature subset F contains to facilitate classification. Then, for each candidate feature subset, its information gain may be calculated based on the training data set D as a candidate feature subset evaluation criterion.

Optionally, the optimal feature subset obtained by the feature selection algorithm is: { game scene ID, mobile phone RAM usage amount, mobile phone RAM free amount, mobile phone CPU usage rate, and mobile phone battery temperature }.

The following describes the training of the decision tree model according to the embodiment of the present invention.

The decision tree of this embodiment is a prediction model, and is also a tree-like decision diagram with an additional probability result, and is used to represent a mapping between object attributes and object values, each node in the tree is used to represent a judgment condition of the object attributes, its branch leaf node is used to represent an object meeting the node condition, and the leaf node of the tree represents a prediction result to which the object belongs. When the client-side jamming problem is solved, the object value in the decision tree model is used for indicating whether jamming occurs or not, the object attribute can be characteristics such as game scene ID, mobile phone RAM usage, mobile phone RAM idle amount, mobile phone CPU usage, mobile phone battery temperature and the like, and the purpose of model training is to obtain the critical judgment condition of the object attribute when the jamming occurs to a player.

The decision tree model of this embodiment comprises a root node, a number of internal nodes and a number of leaf nodes. The leaf nodes correspond to decision results, and each of the other nodes corresponds to an attribute test. The sample set contained by each node is divided into child nodes according to the result of the attribute test. The root node contains a sample corpus and the path from the root node to each leaf node corresponds to a decision test sequence. The decision tree processing aims at generating a tree with strong generalization capability, and the design idea of the basic flow is simple and intuitive specific algorithm flow can be as follows:

Input：train data D＝{(x ₁ ，y ₁ )，(x ₂ ，y ₂ )，…，(x _m ，y _m )}

attribute setA＝{a ₁ ，a ₂ ，…a _d }

Process：function TreeGenerate(D，A)

the key to data processing through decision trees is how to select the optimal partition attributes. Alternatively, as the partitioning process continues, it is desirable that the samples contained in the branch nodes of the decision tree belong to the same class as much as possible, i.e., the "purity" of the nodes is higher and higher. This example chooses to use the Gini Index (Gini Index) as the attribute partitioning criterion, and the purity of the data set D can be measured in terms of the Gini value:

wherein p is _k And k =1,2, \ 8230λ, representing the proportion of the kth type samples in the current sample set D. p is a radical of formula _k 'is used to indicate the proportion of the kth' class sample in the current sample set D, and γ is used to indicate the number of classes, for example, 2.

Gini (D) reflects the probability that two samples are randomly drawn from the data set D with inconsistent class labels, so the smaller Gini (D), the higher the purity of the data set D.

For attribute a, its kini index is defined as:

in the candidate attribute set a, the attribute that minimizes the divided kini index is selected as the optimal division attribute, that is:

FIG. 6 is a diagram illustrating the training results of a decision tree model according to an embodiment of the present invention. As shown in fig. 6, a state point class =0 where katton is generated and a state point class =1 where katton is not generated are included. The method comprises the steps of calling a corresponding interface in a decision tree model according to current state data of a client, processing the current state data through an algorithm in the decision tree model, and further obtaining the probability of the client in the stuck state, and when the probability of the client in the stuck state is higher than a certain threshold value, taking optimization measures for the client to avoid the client from being stuck state, and further improving the operation efficiency of the client.

It should be noted that the training data shown in fig. 6 is only an example and does not limit the embodiment of the present invention.

FIG. 7 is a schematic diagram of an interface for optimizing a game client according to an embodiment of the present invention. As shown in fig. 7, the game client is a game client for a personal race, and when the probability that the game client is in a stuck state is predicted to be higher than a certain threshold, a prompt message is displayed, for example, "it is detected that the fluency of the last game is too low," it is strongly recommended that the image quality be reduced to improve the fluency. Is fluency reduced? ". When the user selects the 'confirm' button, the fluency of the game client is reduced so as to reduce the probability of the game client appearing in a stuck state. When the user selects the "cancel" button, the game client may be stuck in the next time as the smoothness of the game client is not reduced by the client.

In this embodiment, status data of the client is obtained, where the status data includes RAM usage of the terminal, CPU usage of the terminal, battery temperature of the terminal, scenario ID of the client, FPS of the client, and the like. And calculating the historical stuck condition of the client through a stuck calculation formula based on the FPS, and quantifying the stuck of the client to be used as the marking data of model training. And then, carrying out feature selection on the marked data, selecting features with higher correlation, and finally training a decision tree model by using the features with higher correlation. And then, in the running process of the client, calculating the probability of the client jamming in the running process in real time according to the trained model. When the probability of occurrence of the jamming is larger than or equal to a certain threshold value, an optimization strategy is immediately adopted for the client, so that the jamming of the client is prevented, the operation efficiency of the client is improved, and the user experience is further improved.

According to another aspect of the embodiment of the present invention, a data processing apparatus for implementing the data processing method is also provided. Fig. 8 is a schematic diagram of a data processing apparatus according to an embodiment of the present invention. As shown in fig. 8, the apparatus may include: a determination unit 10 and a processing unit 20.

The determining unit 10 is configured to determine, through the target model and current state data of the client, a target probability that a stuck state occurs during a process in which the client operates in a first time period, where the current state data includes state data generated by the client and/or a terminal in which the client is installed during a process in which the client operates in a current second time period, and the second time period is earlier than the first time period.

And the processing unit 20 is configured to, if the target probability is greater than or equal to the first target threshold, process current state data of the client, and/or perform a target operation on at least one of the client and the terminal, so that a probability that a stuck state occurs in a process in which the client operates over the first time period is lower than the target probability.

Optionally, the apparatus further comprises: an acquisition unit and a training unit. The client-side state data acquiring unit is used for acquiring target state data before determining a target probability of a stuck state occurring in the process of the client-side running in a first time period through a target model and current state data of the client-side, wherein the target state data comprises state data generated by the client-side and/or a terminal for installing the client-side under the condition that the stuck state occurs in the process of the client-side running in a past third time period, and the third time period is earlier than a second time period; and the training unit is used for training the initial training model by using the target state data to obtain the trained target model.

Optionally, the obtaining unit includes: the device comprises an acquisition module and a marking module. The acquisition module is used for acquiring historical state data, wherein the historical state data comprises state data generated by the client and/or a terminal for installing the client in the process that the client runs on a third past time period; and the marking module is used for marking the target state data under the target frame rate in the historical state data, wherein the target frame rate is the frame rate of the client when the client is in the stuck state.

Optionally, the obtaining unit further comprises: the selection module is used for selecting first target characteristic data from the target state data after the target state data at the target frame rate are marked, wherein the first target characteristic data are state data, of which the correlation degree between the historical state data and the pause state of the client is higher than a second target threshold value; the training unit comprises: and the first training module is used for training the initial training model by using the first target characteristic data in the target state data to obtain the trained target model.

Optionally, the selection module comprises: the device comprises a selection sub-module, an evaluation sub-module, a first acquisition sub-module and a determination sub-module. The selection submodule is used for selecting candidate feature data from the target state data, wherein the number of the candidate feature data is a first number; the evaluation sub-module is used for evaluating the candidate feature data of the first quantity respectively to obtain evaluation results of the first quantity; the first obtaining submodule is used for obtaining a target evaluation result which meets a target condition in the first number of evaluation results; and the determining submodule is used for determining the candidate characteristic data corresponding to the target evaluation result as first target characteristic data.

Optionally, the evaluation sub-module obtains a first number of evaluation results by performing the following steps to respectively evaluate a first number of candidate feature data: respectively obtaining information entropy gains of a first number of candidate characteristic data to obtain a first number of information entropy gains, wherein the evaluation result comprises the information entropy gains; the first obtaining sub-module obtains a target evaluation result meeting the target condition from the first number of evaluation results by performing the following steps: obtaining information entropy gains greater than a third target threshold from the first number of information entropy gains; and determining the information entropy gain larger than the third target threshold value as a target evaluation result.

Optionally, the marking module comprises: the system comprises a second obtaining submodule, a third obtaining submodule, a fourth obtaining submodule, a fifth obtaining submodule and a marking submodule. The second obtaining submodule is used for obtaining a plurality of frame rates generated by the client in the target time period; the third obtaining submodule is used for obtaining a first weight corresponding to an average frame rate of the plurality of frame rates, wherein the average frame rate is in direct proportion to the first weight, and the first weight is used for indicating the influence proportion of the average frame rate on the pause state of the client; the fourth obtaining submodule is used for obtaining a second weight corresponding to the variance of the plurality of frame rates, wherein the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the pause state of the client; a fifth obtaining submodule, configured to obtain a target score through the first weight and the second weight, where the target score is proportional to an average frame rate of the plurality of frame rates and inversely proportional to a variance of the plurality of frame rates; and the marking submodule is used for determining the frame rate corresponding to the target score as the target frame rate under the condition that the target score is smaller than the fourth target threshold, and marking the target state data under the target frame rate in the historical state data.

Optionally, the fourth obtaining sub-module is configured to obtain the second weight corresponding to the variance of the plurality of frame rates by: dividing a plurality of frame rates into a plurality of frame rate groups; acquiring the variance of the frame rate in each group of frame rate groups; acquiring a target frame rate group corresponding to the variance larger than a fourth target threshold; and acquiring a second weight corresponding to the variance of the frame rates in the target frame rate group.

Optionally, the training unit comprises: the device comprises a first determining module, a second determining module and a second training module. The first determining module is used for determining the target state data as the object attribute of the target model; the second determining module is used for determining the result of the client terminal in the stuck state as an object value of the target model, wherein the object attribute and the object value have a mapping relation in the target model; and the second training module is used for training the initial training model through the object attributes and the object values to obtain a trained target model, and the target state data in the trained target model is critical data for determining that the client has a stuck state.

The determination unit 10 includes: and the third determining module is used for determining the target probability of the pause state occurring in the process of the client operating in the first time period through the target model and the second target state data, wherein the current state data of the client comprises the second target state data, and the second target characteristic data is the state data of which the correlation degree between the current state data and the pause state of the client is higher than a second target threshold value.

Optionally, the apparatus further comprises: the device comprises a first acquisition unit and a first training unit. The first acquiring unit is used for acquiring the state data of the client in the actual running process at target time intervals after processing the current state data of the client and/or performing target operation on at least one of the client and the terminal; and the first training unit is used for training the target model through the state data of the client in the actual operation process so as to update the target model.

It should be noted that the determining unit 10 in this embodiment may be configured to execute step S202 in this embodiment, and the processing unit 20 in this embodiment may be configured to execute step S204 in this embodiment.

In the embodiment, the determining unit 10 determines a target probability of a stuck state occurring during the operation of the client over a first time period through a target model and current state data of the client, where the current state data includes state data generated by the client and/or a terminal on which the client is installed during the operation of the client over a current second time period, and the processing unit 20 processes the current state data of the client and/or performs a target operation on at least one of the client and the terminal under the condition that the target probability is greater than or equal to a first target threshold value, so that the probability of the stuck state occurring during the operation of the client over the first time period is lower than the target probability. The target probability of the stuck state occurring in the process of the client operating in the first time period is determined through the target model and the current state data of the client, so that the aim of reducing the probability of the stuck state occurring in the client by taking effective measures is fulfilled, the problem that the stuck state of the client cannot be avoided by taking effective measures is avoided, the technical effect of improving the operating efficiency of the client is achieved, and the technical problem of low operating efficiency of the client in the related technology is solved.

It should be noted that, the above units and modules are the same as the examples and application scenarios realized by the corresponding steps, but are not limited to the disclosure of the above embodiments. It should be noted that the modules described above as a part of the apparatus may be operated in a hardware environment as shown in fig. 1, and may be implemented by software, or may be implemented by hardware, where the hardware environment includes a network environment.

According to another aspect of the embodiments of the present invention, there is also provided an electronic device for implementing the data processing method.

Fig. 9 is a block diagram of an electronic device according to an embodiment of the present invention. As shown in fig. 9, the electronic device may include: comprising a memory 901 and a processor 903, the memory 901 having stored therein a computer program, the processor 903 being arranged to execute the steps of any of the above-described method embodiments by means of the computer program. Optionally, as shown in fig. 8, the electronic apparatus may further include a transmission apparatus 905 and an input-output device 907.

Optionally, in this embodiment, the electronic apparatus may be located in at least one network device of a plurality of network devices of a computer network.

Alternatively, in this embodiment, the processor 903 may be configured to execute the following steps by a computer program:

determining the target probability of the stuck state occurring in the process of the client operating in the first time period according to the target model and the current state data of the client, wherein the current state data comprises state data generated by the client and/or a terminal for installing the client in the process of the client operating in the current second time period, and the second time period is earlier than the first time period;

and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state in the process of the client operating in the first time period is lower than the target probability.

The processor 903 is further configured to perform the steps of: acquiring target state data before determining a target probability of a stuck state occurring in a process of a client operating in a first time period through a target model and current state data of the client, wherein the target state data comprises state data generated by the client and/or a terminal for installing the client under the condition that the stuck state occurs in the process of the client operating in a third time period which is past, and the third time period is earlier than a second time period; training the initial training model by using the target state data to obtain a trained target model;

the processor 903 is further configured to perform the following steps: acquiring historical state data, wherein the historical state data comprises state data generated by the client and/or a terminal for installing the client in the process that the client runs on a third time period which is past; and marking target state data under a target frame rate in the historical state data, wherein the target frame rate is the frame rate of the client when the client is in the pause state.

The processor 903 is further configured to perform the following steps: after acquiring the historical state data, selecting first target characteristic data from the target state data, wherein the first target characteristic data is state data, in the historical state data, of which the degree of correlation with the stuck state of the client is higher than a second target threshold; and training the initial training model by using the first target characteristic data in the target state data to obtain the trained target model.

The processor 903 is further configured to perform the steps of: selecting candidate feature data from the target state data, wherein the number of the candidate feature data is a first number; evaluating the candidate feature data of the first quantity respectively to obtain evaluation results of the first quantity; obtaining target evaluation results meeting the target conditions in the first number of evaluation results; and determining candidate feature data corresponding to the target evaluation result as first target feature data.

The processor 903 is further configured to perform the following steps: respectively obtaining information entropy gains of a first number of candidate characteristic data to obtain a first number of information entropy gains, wherein the evaluation result comprises the information entropy gains; obtaining information entropy gains greater than a third target threshold from the first number of information entropy gains; and determining the information entropy gain larger than the third target threshold value as a target evaluation result.

The processor 903 is further configured to perform the following steps: acquiring a plurality of frame rates generated by a client in a target time period; acquiring a first weight corresponding to an average frame rate of a plurality of frame rates, wherein the average frame rate is in direct proportion to the first weight, and the first weight is used for indicating the influence of the average frame rate on a pause state of a client; acquiring a second weight corresponding to the variance of the plurality of frame rates, wherein the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the stuck state of the client; obtaining a target score through the first weight and the second weight, wherein the target score is in direct proportion to the average frame rate of the plurality of frame rates and in inverse proportion to the variance of the plurality of frame rates; and under the condition that the target score is smaller than a fourth target threshold, determining the frame rate corresponding to the target score as the target frame rate, and marking the target state data at the target frame rate in the historical state data.

The processor 903 is further configured to perform the following steps: dividing a plurality of frame rates into a plurality of frame rate groups; acquiring the variance of the frame rate in each group of frame rate groups; acquiring a target frame rate group corresponding to the variance larger than a fourth target threshold; and acquiring a second weight corresponding to the variance of the frame rates in the target frame rate group.

The processor 903 is further configured to perform the following steps: determining target state data as object attributes of a target model; determining a result of the client in the stuck state as an object value of a target model, wherein the object attribute and the object value have a mapping relation in the target model; and training the initial training model through the object attributes and the object values to obtain a trained target model, wherein the target state data is critical data for determining that the client is in a stuck state in the trained target model.

The processor 903 is further configured to perform the following steps: and determining the target probability of the stagnation state of the client in the process of running in the first time period through the target model and second target state data, wherein the current state data of the client comprises the second target state data, and the second target characteristic data is the state data of which the correlation degree with the stagnation state of the client in the current state data is higher than a second target threshold value.

The processor 903 is further configured to perform the following steps: after processing the current state data of the client and/or performing target operation on at least one of the client and the terminal, acquiring the state data of the client in the actual operation process at target time intervals; and training the target model through the state data of the client in the actual operation process so as to update the target model.

Alternatively, it can be understood by those skilled in the art that the structure shown in fig. 9 is only an illustration, and the electronic device may also be a terminal device such as a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palm computer, and a Mobile Internet Device (MID), PAD, etc. Fig. 9 is a diagram illustrating a structure of the electronic device. For example, the electronic device may also include more or fewer components (e.g., network interfaces, display devices, etc.) than shown in FIG. 9, or have a different configuration than shown in FIG. 9.

The memory 901 may be used to store software programs and modules, such as program instructions/modules corresponding to the data processing method and apparatus in the embodiments of the present invention, and the processor 903 executes various functional applications and data processing by running the software programs and modules stored in the memory 901, that is, implements the data processing method described above. The memory 901 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 901 can further include memory located remotely from the processor 903, which can be connected to a terminal over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission device 905 is used for receiving or transmitting data via a network. Examples of the network may include a wired network and a wireless network. In one example, the transmission device 905 includes a Network adapter (NIC) that can be connected to a router via a Network cable and other Network devices so as to communicate with the internet or a local area Network. In one example, the transmission device 905 is a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

Among them, the memory 901 is used to store an application program in particular.

The embodiment of the invention provides a data processing scheme. Determining the target probability of the client in the stuck state in the process of running in the first time period according to the target model and the current state data of the client; and under the condition that the target probability is greater than or equal to a first target threshold, processing current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of the occurrence of a stuck state of the client in the process of running in a first time period is lower than the target probability. The target probability of the stuck state occurring in the process of the client operating in the first time period is determined through the target model and the current state data of the client, so that the aim of reducing the probability of the stuck state occurring in the client by taking effective measures is fulfilled, the problem that the stuck state of the client cannot be avoided by taking effective measures is avoided, the technical effect of improving the operating efficiency of the client is achieved, and the technical problem of low operating efficiency of the client in the related technology is solved.

Embodiments of the present invention also provide a storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the above method embodiments when executed.

Alternatively, in the present embodiment, the storage medium may be configured to store a computer program for executing the steps of:

Optionally, the storage medium is further arranged to store program code for performing the steps of: acquiring target state data before determining a target probability of a stuck state occurring in a process of a client operating in a first time period through a target model and current state data of the client, wherein the target state data comprises state data generated by the client and/or a terminal for installing the client under the condition that the stuck state occurs in the process of the client operating in a third time period which is past, and the third time period is earlier than a second time period; training the initial training model by using the target state data to obtain a trained target model;

optionally, the storage medium is further arranged to store program code for performing the steps of: acquiring historical state data, wherein the historical state data comprises state data generated by the client and/or a terminal for installing the client in the process that the client runs on a third time period which is past; in the historical state data, target state data at a target frame rate is marked, wherein the target frame rate is the frame rate of the client when the client is in the stuck state.

Optionally, the storage medium is further arranged to store program code for performing the steps of: after acquiring historical state data, selecting first target characteristic data from the target state data, wherein the first target characteristic data is state data of which the degree of correlation with the stuck state of the client in the historical state data is higher than a second target threshold; and training the initial training model by using the first target characteristic data in the target state data to obtain the trained target model.

Optionally, the storage medium is further arranged to store program code for performing the steps of: selecting candidate feature data from the target state data, wherein the number of the candidate feature data is a first number; evaluating the candidate feature data of the first quantity respectively to obtain evaluation results of the first quantity; obtaining target evaluation results meeting target conditions in the first number of evaluation results; and determining candidate feature data corresponding to the target evaluation result as first target feature data.

Optionally, the storage medium is further arranged to store program code for performing the steps of: respectively obtaining information entropy gains of a first number of candidate characteristic data to obtain a first number of information entropy gains, wherein the evaluation result comprises the information entropy gains; the obtaining of the target evaluation result meeting the target condition in the first number of evaluation results includes: obtaining information entropy gains greater than a third target threshold from the first number of information entropy gains; and determining the information entropy gain larger than the third target threshold value as a target evaluation result.

Optionally, the storage medium is further arranged to store program code for performing the steps of: acquiring a plurality of frame rates generated by a client in a target time period; acquiring a first weight corresponding to an average frame rate of a plurality of frame rates, wherein the average frame rate is in direct proportion to the first weight, and the first weight is used for indicating the influence proportion of the average frame rate on the hiton state of the client; acquiring a second weight corresponding to the variance of the plurality of frame rates, wherein the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the hiton state of the client; obtaining a target score through the first weight and the second weight, wherein the target score is in direct proportion to the average frame rate of the plurality of frame rates and in inverse proportion to the variance of the plurality of frame rates; and under the condition that the target score is smaller than a fourth target threshold, determining the frame rate corresponding to the target score as the target frame rate, and marking the target state data at the target frame rate in the historical state data.

Optionally, the storage medium is further arranged to store program code for performing the steps of: dividing a plurality of frame rates into a plurality of frame rate groups; acquiring the variance of the frame rates in each group of frame rate groups; acquiring a target frame rate group corresponding to the variance larger than a fourth target threshold; and acquiring a second weight corresponding to the variance of the frame rates in the target frame rate group.

Optionally, the storage medium is further arranged to store program code for performing the steps of: determining target state data as object attributes of a target model; determining a result of the client in the stuck state as an object value of a target model, wherein the object attribute and the object value have a mapping relation in the target model; and training the initial training model through the object attributes and the object values to obtain a trained target model, wherein the target state data is critical data for determining that the client is in a stuck state in the trained target model.

Optionally, the storage medium is further arranged to store program code for performing the steps of: and determining the target probability of the stuck state of the client in the process of running in the first time period through the target model and the second target state data, wherein the current state data of the client comprises the second target state data, and the second target characteristic data is state data of which the correlation degree with the stuck state of the client in the current state data is higher than a second target threshold value.

Optionally, the storage medium is further arranged to store program code for performing the steps of: after processing the current state data of the client and/or performing target operation on at least one of the client and the terminal, acquiring the state data of the client in the actual operation process at target time intervals; and training the target model through the state data of the client in the actual operation process so as to update the target model.

Optionally, the storage medium is further configured to store a computer program for executing the steps included in the method in the foregoing embodiment, which is not described in detail in this embodiment.

Alternatively, in this embodiment, a person skilled in the art may understand that all or part of the steps in the methods of the foregoing embodiments may be implemented by a program instructing hardware associated with the terminal device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, read-Only memories (ROMs), random Access Memories (RAMs), magnetic or optical disks, and the like.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages and disadvantages of the embodiments.

The integrated unit in the above embodiments, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in the above computer-readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing one or more computer devices (which may be personal computers, servers, network devices, etc.) to execute all or part of the steps of the method according to the embodiments of the present invention.

In the above embodiments of the present invention, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described in detail in a certain embodiment.

In the several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other ways. The above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be implemented in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A data processing method, comprising:

determining a target probability of a stuck state occurring in a process of running the client in a prediction time period according to a target model and current state data of the client, wherein the current state data comprises state data generated by the client and/or a terminal for installing the client in the process of running the client in the current time period, and the current time period is earlier than the prediction time period;

under the condition that the target probability is larger than or equal to a first target threshold, processing the current state data of the client, and/or performing target operation on at least one of the client and the terminal, so that the probability of a stuck state occurring in the process of the client operating on the prediction time period is lower than the target probability;

the determining, by using the target model and the current state data of the client, the target probability of the client in the stuck state in the process of operating in the prediction time period includes:

and determining the target probability of the stuck state occurring in the process of the client operating in the current time period according to the target model and second target state data, wherein the current state data of the client comprises the second target state data, and the second target state data is state data with high correlation degree with the stuck state of the client in the current state data.

2. The method of claim 1, further comprising:

acquiring target state data, wherein the target state data comprises state data generated by the client and/or a terminal for installing the client under the condition that a stuck state occurs in the process that the client runs on a past time period, and the past time period is earlier than the current time period;

and training an initial training model by using the target state data to obtain the trained target model.

3. The method of claim 2, wherein obtaining the target state data comprises:

acquiring historical state data, wherein the historical state data comprises state data generated by the client and/or a terminal installed with the client in the process that the client runs on the past time period;

and marking the target state data under a target frame rate in the historical state data, wherein the target frame rate is the frame rate of the client when the client is in a stuck state.

4. The method of claim 3,

after marking the destination state data at the destination frame rate, the method further comprises: selecting first target characteristic data from the target state data, wherein the first target characteristic data is state data of which the correlation degree between the historical state data and the stuck state of the client is higher than a second target threshold;

training the initial training model by using the target state data to obtain the trained target model, wherein the training comprises: and training the initial training model by using the first target characteristic data in the target state data to obtain the trained target model.

5. The method of claim 4, wherein selecting the first target feature data from the target state data comprises:

selecting candidate feature data from the target state data, wherein the number of the candidate feature data is a first number;

evaluating the candidate feature data of the first quantity respectively to obtain evaluation results of the first quantity;

obtaining target evaluation results meeting target conditions in the first number of evaluation results;

and determining candidate feature data corresponding to the target evaluation result as the first target feature data.

6. The method of claim 5,

evaluating the first number of candidate feature data respectively, and obtaining the first number of evaluation results includes: respectively obtaining information entropy gains of the candidate characteristic data of the first quantity to obtain information entropy gains of the first quantity, wherein the evaluation result comprises the information entropy gains;

acquiring the target evaluation result meeting the target condition in the first number of evaluation results comprises: obtaining the information entropy gains that are greater than a third target threshold among the first number of the information entropy gains; determining the information entropy gain larger than the third target threshold as the target evaluation result.

7. The method of claim 3, wherein marking the target state data at the target frame rate from the historical state data comprises:

acquiring a plurality of frame rates generated by the client in a target time period;

obtaining a first weight corresponding to an average frame rate of the plurality of frame rates, wherein the average frame rate is in direct proportion to the first weight, and the first weight is used for indicating a proportion of influence of the average frame rate on a stuck state of the client;

acquiring a second weight corresponding to the variance of the plurality of frame rates, wherein the second weight is used for indicating the influence proportion of the variance of the plurality of frame rates on the stuck state of the client;

obtaining an object score through the first weight and the second weight, wherein the object score is proportional to an average frame rate of the plurality of frame rates and inversely proportional to a variance of the plurality of frame rates;

and under the condition that the target score is smaller than a fourth target threshold, determining the frame rate corresponding to the target score as the target frame rate, and marking the target state data at the target frame rate in the historical state data.

8. The method of claim 7, wherein obtaining the second weight corresponding to the variance of the plurality of frame rates comprises:

dividing the plurality of frame rates into a plurality of frame rate groups;

acquiring the variance of the frame rate in each group of frame rate groups;

acquiring a target frame rate group corresponding to the variance larger than a fourth target threshold;

and acquiring the second weight corresponding to the variance of the frame rates in the target frame rate group.

9. The method of claim 2, wherein training the initial training model using the target state data, resulting in a trained target model comprises:

determining the target state data as object attributes of the target model;

determining a result of the client terminal appearing in the stuck state as an object value of the target model, wherein the object attribute and the object value have a mapping relation in the target model;

and training the initial training model according to the object attributes and the object values to obtain the trained target model, wherein the target state data is critical data for determining the occurrence of the stuck state of the client in the trained target model.

10. The method according to any of claims 1 to 9, wherein after processing the current state data of the client and/or performing the target operation on at least one of the client and the terminal, the method further comprises:

acquiring state data of the client in the actual operation process at intervals of target time;

and training the target model through the state data of the client in the actual operation process so as to update the target model.

11. A data processing apparatus, comprising:

the device comprises a determining unit, a judging unit and a judging unit, wherein the determining unit is used for determining the target probability of a stuck state occurring in the process of running the client on a prediction time period through the current state data of the client, the current state data comprises the state data generated by the client and/or a terminal for installing the client in the process of running the client on the current time period, and the current time period is earlier than the prediction time period;

the processing unit is used for processing the current state data of the client under the condition that the target probability is greater than or equal to a first target threshold, and determining the target probability of a stuck state occurring in the process of the client operating in a prediction time period according to the current state data of the client so that the probability of the stuck state occurring in the process of the client operating in the prediction time period is lower than the target probability;

the determination unit includes: a third determining module, configured to determine, through the target model and second target state data, the target probability that the stuck state occurs during a process in which the client operates in the current time period, where the current state data of the client includes the second target state data, and the second target state data is state data with a high degree of correlation between the current state data and the stuck state of the client.

12. The apparatus of claim 11, further comprising:

an acquisition unit configured to acquire target status data, wherein the target status data includes status data generated by the client and/or a terminal that installs the client in a case where a stuck status occurs during the client operates over a past period of time that has passed earlier than the current period of time;

and the training unit is used for training an initial training model by using the target state data to obtain the trained target model.

13. A storage medium, in which a computer program is stored, wherein the computer program is arranged to perform the data processing method of any of claims 1 to 10 when executed.

14. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to execute the data processing method of any of claims 1 to 10 by means of the computer program.