Game test optimization method and system based on machine learning

Abstract

The invention discloses a game test optimization method and system based on machine learning, and relates to the technical field of game test optimization. The method comprises the following steps: acquiring test data, wherein the test data comprises game balance data and informal playing method data; preprocessing the test data, and constructing confrontation balance features and off-track behavior features according to the preprocessed test data; calculating to obtain a balance score according to the confrontation balance characteristics, and calculating to obtain a comprehensive abnormal behavior score according to the track crossing behavior characteristics; and performing threshold judgment on the balance score, generating a first optimization signal when a threshold condition is exceeded, performing threshold judgment on the comprehensive abnormal behavior score, and generating a second optimization signal when the threshold condition is exceeded. Through an analysis method based on machine learning, game balance optimization, abnormal behavior automatic detection and data-driven decision support are realized, and fairness, safety and development efficiency of the game are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of game test optimization, and specifically to a game test optimization method and system based on machine learning. Background Art

[0002] With the rapid development of the game industry, the balance of games and the monitoring and optimization of player behavior have become important links in game design and testing. Game balance refers to the fairness among various characters, weapons, skills, and confrontation combinations in the game, ensuring that no party has an excessive advantage in the game. Moreover, player behavior, especially abnormal behavior such as cheating and exploiting system vulnerabilities, may also affect the game experience and fairness. Therefore, developing an effective game test optimization method that can monitor, evaluate the game in real time during the development and operation stages, and automatically optimize according to data feedback has become the key to improving game quality and player experience.

[0003] Traditional game test and optimization methods mainly rely on manual evaluation and traditional rules, but these methods are difficult to efficiently handle the complex interaction mechanisms of games and the massive player behavior data. To make up for this deficiency, machine learning-based methods have gradually become the mainstream. It can analyze a large amount of game data, automatically learn and identify balance problems and abnormal player behaviors in the game, so as to achieve real-time game optimization.

[0004] The present invention proposes a game test optimization method based on machine learning. Through automated data analysis, it constructs confrontation balance features and deviant behavior features, combines balance scores and abnormal behavior scores, generates optimization signals, discovers problems in the game in a timely manner and conducts optimization, strengthens the balance of the game, quickly identifies and responds to abnormalities in player behavior, and improves the efficiency and quality of game testing. Summary of the Invention

[0005] Based on the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a game test optimization method and system based on machine learning to solve the above technical problems.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: A game test optimization method based on machine learning, including:

[0007] Obtain test data, where the test data includes game balance data and informal gameplay data;

[0008] Preprocess the test data, and construct confrontation balance features and deviant behavior features according to the preprocessed test data;

[0009] Calculate a balance score according to the confrontation balance features, and calculate a comprehensive abnormal behavior score according to the deviant behavior features;

[0010] Perform a threshold judgment on the balance score. When the threshold condition is exceeded, generate a first optimization signal. Perform a threshold judgment on the comprehensive abnormal behavior score. When the threshold condition is exceeded, generate a second optimization signal.

[0011] The present invention is further configured such that the game balance data includes the confrontation combinations used by players, the win-loss records of player confrontations, the combination usage frequencies, the combination battle data, and the confrontation combination effects, and the informal gameplay data includes the player activity duration, the player activity level, the number of abnormal behaviors, and the abnormal behavior frequency; the confrontation balance features include the combination usage frequency, the relative strength of the combination, and the combination effect difference, and the deviant behavior features include the informal behavior frequency and type, the behavior abnormality degree, and the system impact of the abnormal behavior.

[0012] The present invention is further configured such that the combination usage frequency includes the usage frequency and the win rate difference;

[0013] The relative strength of the combination includes the relative strength and the confrontation balance degree;

[0014] The combination effect difference includes the effect deviation and the combination diversity;

[0015] The balance score is calculated based on the usage frequency, the win rate difference, the relative strength, the confrontation balance degree, the effect deviation, and the combination diversity.

[0016] The present invention is further configured such that the calculation logic of the usage frequency is: where F use is the usage frequency, is the usage frequency of the i-th character, weapon, and skill combination, is the player activity level, and λ1 is the influence coefficient for adjusting the impact between the usage frequency and the player activity level;

[0017] The calculation logic of the win rate difference is: where D win is the win rate difference, is the win rate of the i-th combination, W average is the average win rate of all combinations, and λ2 is the influence coefficient between the win rate and the usage frequency;

[0018] The calculation logic of the relative strength is: where R strength is the relative strength, is the number of times the i-th element participates in the confrontation, and γ1 is the influence coefficient for adjusting the impact between the strength and the confrontation;

[0019] The calculation logic of the confrontation balance degree is: where B combat is the confrontation balance degree, is the number of confrontations for the i-th combination, is the failure rate during the confrontation of the i-th combination, and γ2 is the influence coefficient between the confrontation balance and the player activity;

[0020] The calculation logic of the said effect deviation is: where E bias is the effect deviation, is the effect value of the i-th combination of character, weapon and skill, and E average is the average value of the effect values of all combinations, is the usage frequency of the i-th combination, and δ1 is the influence coefficient between the effect deviation and the usage frequency;

[0021] The calculation logic of the said combination diversity is where T max is the maximum number of uses of all combinations in the system, and δ2 is the influence coefficient between the effect and the usage frequency;

[0022] The calculation logic of the said balance score is: where B score is the balance score, is the usage frequency of the i-th character, weapon or skill, is the winning rate difference of the i-th combination, is the relative strength of the i-th combination, is the confrontation balance of the i-th combination, is the effect deviation of the i-th combination, is the combination diversity of the i-th combination, and w1, w2, w3, w4, w5 and w6 are weight coefficients.

[0023] The present invention is further configured that the informal behavior frequency and type include the abnormal behavior frequency and the abnormal behavior type;

[0024] The said behavior abnormality includes the behavior abnormality degree and the behavior deviation degree;

[0025] The system influence of the said abnormal behavior includes the system influence degree and the vulnerability exploitation influence degree;

[0026] The comprehensive abnormal behavior score is calculated according to the said abnormal behavior frequency, the said abnormal behavior type, the said behavior abnormality degree, the said behavior deviation degree, the said system influence degree and the said vulnerability exploitation influence degree.

[0027] The present invention is further configured that the calculation logic of the said abnormal behavior frequency is: where F anomaly is the abnormal behavior frequency, is the number of the i-th abnormal behavior, For the player's game duration, α1 is the decay coefficient between abnormal behavior and session duration;

[0028] The calculation logic of the abnormal behavior type is: Among them, Tanomaly is the abnormal behavior type, βi is the weight coefficient of behavior type i, is the severity score of the i-th behavior, is the player's behavior pattern, and α2 is the relationship coefficient between the player's behavior pattern and the abnormal behavior type;

[0029] The calculation logic of the behavior abnormality degree is: Among them, A anomaly is the behavior abnormality degree, is the complexity score of the i-th abnormal behavior, is the time point when the i-th abnormal behavior occurs, β1 is the influence coefficient between the abnormal behavior time point and the complexity, is the complexity score of the i-th normal behavior;

[0030] The calculation logic of the behavior deviation degree is: Among them, P dewi2tion is the behavior deviation degree, is the deviation degree between the i-th abnormal behavior and the normal behavior, is the complexity of the task when the i-th abnormal behavior occurs, is the measurement value of the i-th normal behavior, and κ1 is the coefficient between the task complexity and the behavior deviation degree;

[0031] The calculation logic of the system influence degree is: Among them, I impact is the system influence degree, is the influence score of the i-th abnormal behavior on the game system, is the occurrence frequency of the i-th abnormal behavior, is the occurrence frequency of the i-th normal behavior, and μ1 is the system influence degree coefficient;

[0032] The calculation logic of the vulnerability exploitation influence degree is: Among them, V exploit is the vulnerability exploitation influence degree, is the severity score of the i-th vulnerability exploitation, is the player activity of the i-th vulnerability exploitation, is the score of the i-th normal behavior, and μ2 is the coefficient between the vulnerability exploitation influence degree and the activity;

[0033] The calculation logic of the comprehensive abnormal behavior score is: Among them, S anomaly is the comprehensive abnormal behavior score, is the frequency of the i-th abnormal behavior, is the type and severity of the i-th abnormal behavior, is the abnormality degree of the i-th abnormal behavior, is the deviation degree of the i-th behavior, is the impact degree of the i-th abnormal behavior on the system, is the impact degree of the i-th vulnerability exploitation, and ρ1, ρ2, ρ3, ρ4, ρ5 and ρ6 are weight coefficients.

[0034] The present invention is further configured to perform a threshold judgment on the balance score. When the balance score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated;

[0035] The setting standard of the upper threshold is: the mean value of the balance score plus the standard deviation of the balance score;

[0036] The setting standard of the lower threshold is: the mean value of the balance score minus the standard deviation of the balance score;

[0037] The calculation logic of the mean value of the balance score is: where avg(B score ) is the mean value of the balance score in the game within a period of time, is the balance score calculated for the i-th time;

[0038] The calculation logic of the standard deviation of the balance score is: where std(B score ) is the standard deviation of the balance score in the game within a period of time.

[0039] The present invention is further configured to perform a threshold judgment on the comprehensive abnormal behavior score. When the comprehensive abnormal behavior score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated;

[0040] The setting standard of the upper threshold is: the mean value of the comprehensive abnormal behavior score plus the standard deviation of the comprehensive abnormal behavior score;

[0041] The setting standard of the lower threshold is: the mean value of the comprehensive abnormal behavior score minus the standard deviation of the comprehensive abnormal behavior score;

[0042] The calculation logic of the mean value of the comprehensive abnormal behavior score is: where avg(S anomaly ) is the mean value of the comprehensive abnormal behavior score in the game within a period of time, is the comprehensive abnormal behavior score calculated for the i-th time;

[0043] The calculation logic of the standard deviation of the comprehensive abnormal behavior score is: where std(Sanomaly ) is the standard deviation of the comprehensive abnormal score in the game over a period of time.

[0044] The present invention is further configured to obtain the above-mentioned balance score and comprehensive abnormal behavior score in real time through machine learning for visual display, and give an early warning when the balance score and comprehensive abnormal behavior score do not meet the threshold conditions; give an early warning when the mutation value is greater than the preset mutation value threshold, where the mutation value is the slope of the balance score and comprehensive abnormal behavior score.

[0045] The present invention also provides a game test optimization system based on machine learning, and the system includes:

[0046] Data acquisition module: used to acquire test data, and the test data includes game balance data and informal gameplay data;

[0047] Preprocessing and construction module: used to preprocess the test data and construct anti-balance features and deviant behavior features according to the preprocessed test data;

[0048] Calculation module: used to calculate the balance score according to the anti-balance features and calculate the comprehensive abnormal behavior score according to the deviant behavior features;

[0049] Threshold judgment module: used to judge the threshold of the balance score, and generate a first optimization signal when the threshold condition is exceeded, and judge the threshold of the comprehensive abnormal behavior score, and generate a second optimization signal when the threshold condition is exceeded.

[0050] The present invention provides a game test optimization method and system based on machine learning. The method includes acquiring test data, where the test data includes game balance data and informal gameplay data; preprocessing the test data and constructing anti-balance features and deviant behavior features according to the preprocessed test data; calculating the balance score according to the anti-balance features and calculating the comprehensive abnormal behavior score according to the deviant behavior features; judging the threshold of the balance score, and generating a first optimization signal when the threshold condition is exceeded, and judging the threshold of the comprehensive abnormal behavior score, and generating a second optimization signal when the threshold condition is exceeded. The beneficial effects produced include:

[0051] 1. Improve game balance: Through the machine learning-based analysis method, it is possible to calculate and evaluate the balance score in the game in real time, helping developers discover potential imbalance situations. Through a comprehensive analysis of data such as player behavior, confrontation combinations, and win rate differences, ensure that all elements in the game are in a state of fair competition and avoid the appearance of overly powerful characters or combinations;

[0052] 2. Automated detection of abnormal behavior: It can automatically identify the abnormal behavior of players and calculate a comprehensive abnormal behavior score. By analyzing dimensions such as the frequency, type, abnormality degree of abnormal behavior and its impact on the system, the system can timely detect and feedback abnormal behavior, so as to quickly respond and improve game security and player experience;

[0053] 3. Data-driven decision support: Provide data-based decision support tools. Developers and testers can rely on the analysis results automatically generated by the system to quickly make adjustment decisions without relying on subjective human judgment, making game development and testing more scientific and accurate.

[0054] The above description is only an overview of the technical solution of this application. In order to be able to more clearly understand the technical means of this application, it can be implemented in accordance with the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically gives the specific implementation manners of this application. Brief Description of the Drawings

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings. In the drawings:

[0056] Figure 1 It is a flowchart of a game test optimization method based on machine learning shown in an exemplary embodiment of the present invention;

[0057] Figure 2 It is a schematic structural diagram of a game test optimization system based on machine learning shown in an exemplary embodiment of the present invention. Detailed Description of the Invention

[0058] The following will describe the embodiments of the present invention with reference to the drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for explaining the present invention, rather than for limiting the protection scope of the present invention.

[0059] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present invention. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0060] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0061] Embodiment 1

[0062] A game testing optimization method based on machine learning, as Figure 1 shown, includes:

[0063] Obtain test data, where the test data includes game balance data and informal gameplay data;

[0064] Preprocess the test data, and construct an adversarial balance feature and a deviant behavior feature based on the preprocessed test data;

[0065] Calculate a balance score based on the adversarial balance feature, and calculate a comprehensive abnormal behavior score based on the deviant behavior feature;

[0066] Perform a threshold judgment on the balance score. When the threshold condition is exceeded, generate a first optimization signal. Perform a threshold judgment on the comprehensive abnormal behavior score. When the threshold condition is exceeded, generate a second optimization signal.

[0067] Specifically, the game balance data includes the adversarial combinations used by players, the win-loss records of player battles, the combination usage frequencies, the combination battle data, and the adversarial combination effects. The informal gameplay data includes the player activity duration, player activity level, the number of abnormal behaviors, and the abnormal behavior frequency. The adversarial balance features include the combination usage frequency, the relative strength of the combination, and the combination effect difference. The deviant behavior features include the informal behavior frequency and type, the behavior abnormality degree, and the system impact of the abnormal behavior.

[0068] The present invention is further configured such that the combination usage frequency includes the usage frequency and the win rate difference. The calculation logic of the usage frequency is: Where F use is the usage frequency, is the usage frequency of the i-th character, weapon, and skill combination, Let \(A\) be the player activity level, and \(\lambda_1\) be the influence coefficient that adjusts the relationship between the usage frequency and the player activity level. The calculation logic for the win rate difference is as follows: where \(D\) win is the win rate difference, is the win rate of the \(i\)-th combination, \(W\) average is the average win rate of all combinations, and \(\lambda_2\) is the influence coefficient between the win rate and the usage frequency. Specifically, \(F\) use reflects the usage frequency of the \(i\)-th character, weapon, and skill combination in the game by attenuating the relationship between the player activity level and the frequency; \(\lambda_1\) is used to measure the influence degree of the player activity level on the usage frequency, and its value range is \((0, 10]\); \(D\) win reflects the difference degree between the win rate of the \(i\)-th combination and the win rates of other combinations by considering the relationship between the combination usage frequency and the win rate. A larger difference indicates that the \(i\)-th combination may be unbalanced and needs to be optimized; \(\lambda_2\) is used to measure the influence of the win rate difference on the combination usage frequency, and its value range is \((0, 5]\); By calculating and combining the player activity level and the combination win rate, a data-driven method is provided to automatically detect and adjust the balance in the game, making the game development more scientific and precise;

[0069] The relative strength of the combination includes relative strength and confrontation balance degree. The calculation logic for the relative strength is as follows: where \(R\) strength is the relative strength, is the number of times the \(i\)-th element participates in the confrontation, and \(\gamma_1\) is the influence coefficient that adjusts the relationship between the strength and the confrontation; The calculation logic for the confrontation balance degree is as follows: where \(B\) combat is the confrontation balance degree, is the number of confrontations of the \(i\)-th combination, is the failure rate when the \(i\)-th combination confronts, and \(\gamma_2\) is the influence coefficient between the confrontation balance degree and the player activity level. Specifically, \(R\) strength is used to measure the performance of the \(i\)-th combination in the confrontation with other combinations; \(\gamma_1\) is used to adjust the attenuation degree of the influence of the win rate on the relative strength when the number of confrontations increases, and its value range is \((0, 5]\); \(B\) combat is used to measure the confrontation balance state between different combinations in the game; \(\gamma_2\) is used to adjust the influence degree of the player activity level on the confrontation balance degree, and its value range is \((0, 5]\); By calculating the relative strength and the confrontation balance degree, the characters or combinations that dominate in the confrontation can be identified, and balance adjustments can be made to ensure that the competitiveness of different combinations in the game is relatively fair;

[0070] The combination effect difference includes effect deviation and combination diversity. The calculation logic for the effect deviation is as follows: where \(E\) bias is the effect deviation, is the effect value E for the i-th combination of character, weapon and skill average is the average effect value of all combinations is the usage frequency of the i-th combination, and δ1 is the influence coefficient between the effect deviation and the usage frequency; the calculation logic of the combination diversity is as follows where T max is the maximum usage times of all combinations in the system, and δ2 is the influence coefficient between the effect and the usage frequency; specifically, E bias is used to measure the difference between the effect of each combination of character, weapon and skill and the average effect value. When the effect deviation value is larger, it means that some combinations are too powerful or too weak, affecting the game balance; δ1 is used to control the influence degree of the usage frequency on the effect deviation, and the value range is (0, 5]; V combo is used to measure the diversity of all combinations of characters, weapons and skills in the game. When the diversity is low, it means that some combinations are too powerful, resulting in a single game experience and need to be optimized; δ2 is used to control the influence degree of the combination effect value on the diversity, and the value range is (0, 5]; Based on the automatic calculation results of the effect deviation and the diversity, the game design can be optimized more precisely, avoiding unfair games caused by overly powerful combinations, reducing overly single or repetitive game experiences, and enhancing the game fun of players;

[0071] Calculate the balance score according to the usage frequency, the win rate difference, the relative strength, the confrontation balance degree, the effect deviation and the combination diversity; the calculation logic of the balance score is as follows where B score is the balance score is the usage frequency of the i-th character, weapon or skill is the win rate difference of the i-th combination is the relative strength of the i-th combination is the confrontation balance degree of the i-th combination is the effect deviation of the i-th combination is the combination diversity of the i-th combination, and w1, w2, w3, w4, w5 and w6 are weight coefficients; specifically, B score Quantify the balance in a game system by comprehensively considering the usage frequency, the win rate difference, the relative strength, the confrontation balance degree, the effect deviation and the combination diversity is used to reflect the frequency of players choosing the i-th combination is used to measure the relative gap between the win rate of the i-th combination and other combinations is for the actual combat performance of the i-th combination is used to evaluate the performance balance of the i-th combination in game confrontation Used to measure the degree to which the performance of the i-th combination deviates from the average level; Used to reflect the diversity degree of the i-th combination in the entire game environment; w1, w2, w3, w4, w5, and w6 are used to represent the contribution degree of the corresponding indicators to the final balance score, and the value range is [0, 1]; By comprehensively evaluating multiple dimensions, the balance in the game can be quantified, helping to identify potential unbalanced combinations.

[0072] The present invention is further configured such that the informal behavior frequency and type include abnormal behavior frequency and abnormal behavior type; The present invention is further configured such that the calculation logic of the abnormal behavior frequency is: Where, F anomaly Is the abnormal behavior frequency,

[0073] Is the number of abnormal behaviors in the i-th time, Is the game duration of the player, α1 is the decay coefficient between the abnormal behavior and the conversation duration; The calculation logic of the abnormal behavior type is: Where, T anomaly Is the abnormal behavior type, β i Is the weight coefficient of behavior type i, Is the severity score of the i-th behavior, Is the behavior pattern of the player, α2 is the relationship coefficient between the player behavior pattern and the abnormal behavior type; Specifically, F anomaly The frequency of abnormal behavior of the player within a period of time is obtained through the number of abnormal behaviors and the conversation duration in each player's game session; α1 is used to represent the decay degree of abnormal behavior with the game duration, and the value range is [0.01, 0.5]; T anomaly Is used to measure the influence of different types of abnormal behaviors on the player's behavior pattern; β i Is used to represent the influence degree of each type of abnormal behavior on the game, and the value range is [0, 1]; α2 is used to describe the influence of the player behavior pattern on the abnormal behavior type, and the value range is [0.01, 0.5]; By detecting abnormal behaviors and providing real-time feedback, the game design and adjustment mechanism can be optimized targeted, improving the satisfaction and participation of players;

[0074] The behavior abnormality degree includes behavior abnormality degree and behavior deviation degree; The calculation logic of the behavior abnormality degree is: Where, A anomaly Is the behavior abnormality degree, Is the complexity score of the i-th abnormal behavior, Is the time point when the i-th abnormal behavior occurs, κ1 is the influence coefficient between the abnormal behavior time point and the complexity, is the complexity score of the i-th normal behavior; the calculation logic of the behavior deviation degree is as follows: where P deviatiob is the behavior deviation degree, is the deviation degree between the i-th abnormal behavior and the normal behavior,

[0075] is the complexity of the task when the i-th abnormal behavior occurs, is the measurement value of the i-th normal behavior; κ2 is the coefficient between the task complexity and the behavior deviation degree; specifically, A 3nomaly is used to reflect the difference degree between the player's behavior and the normal behavior; κ1 is used to control the influence of the time distance on the abnormality calculation, and its value range is [0.1, 1]; P deviation is used to evaluate the gap of the abnormal behavior relative to the normal behavior; κ2 is used to control the influence of the task complexity on the deviation degree, and its value range is [0.5, 3]; by evaluating the complexity of the behavior and the time factor, and analyzing the abnormal behavior and the task complexity, early or complex abnormal behaviors can be effectively detected, and potential improper behaviors can be identified;

[0076] The system impacts of the abnormal behavior include the system impact degree and the vulnerability exploitation impact degree; the calculation logic of the system impact degree is as follows: where I impact is the system impact degree, is the impact score of the i-th abnormal behavior on the game system, is the occurrence frequency of the i-th abnormal behavior, is the occurrence frequency of the i-th normal behavior; μ1 is the system impact degree coefficient; the calculation logic of the vulnerability exploitation impact degree is as follows: where V exploit is the vulnerability exploitation impact degree, is the severity score of the i-th vulnerability exploitation, is the player activity of the i-th vulnerability exploitation, is the score of the i-th normal behavior; μ2 is the coefficient between the vulnerability exploitation impact degree and the activity; specifically, I impact is used to measure the impact of the abnormal behavior on the system; μ1 is used to control the attenuation rate of the normal behavior frequency on the system impact degree, and its value range is [0, 5]; V exploit is used to measure the influence degree of the vulnerability exploitation on the game system; μ2 is used for the attenuation speed between the vulnerability exploitation and the activity, and its value range is [0, 5]; by calculating the impact of each abnormal behavior and vulnerability exploitation on the system, more reasonable measures can be taken to maintain the fairness and stability of the game;

[0077] Calculate a comprehensive abnormal behavior score based on the abnormal behavior frequency, the abnormal behavior type, the behavior abnormality degree, the behavior deviation degree, the system impact degree, and the vulnerability exploitation impact degree; the calculation logic of the comprehensive abnormal behavior score is as follows: where S anomaly is the comprehensive abnormal behavior score, is the abnormal behavior frequency of the i-th time, is the type and severity of the abnormal behavior of the i-th time, is the abnormality degree of the abnormal behavior of the i-th time, is the behavior deviation degree of the i-th time, is the system impact degree of the abnormal behavior of the i-th time, is the vulnerability exploitation impact degree of the i-th time, and ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 are weight coefficients; specifically, S abomaly Comprehensively evaluate the severity of abnormal behavior through multiple dimensions of the frequency, type, abnormality degree, deviation degree, system impact degree, and vulnerability exploitation impact degree of abnormal behavior, which is used to measure the abnormality degree of player behavior and its potential impact on the game system Used to represent the occurrence frequency of the i-th type of abnormal behavior, including cheating, hanging up, malicious operations Used to represent the type and severity of the i-th type of abnormal behavior Used for the abnormality degree of the i-th type of abnormal behavior relative to normal behavior; Used to represent the deviation degree of the i-th type of abnormal behavior from normal behavior; Used to measure the degree of damage or influence range of the i-th type of abnormal behavior on the system; Used to measure the degree of damage to the system during vulnerability exploitation; ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 are used to adjust the importance of each index of frequency, type, abnormality degree, deviation degree, system impact degree, and vulnerability exploitation impact degree in the score, and the value range is [0, 1]; through the change of the comprehensive abnormal behavior score, game developers can dynamically adjust the game mechanism to reduce the abused vulnerabilities or unfair behaviors.

[0078] The present invention is further configured to perform a threshold judgment on the balance score. When the balance score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated; the setting standard of the upper threshold is: the balance score mean plus the balance score standard deviation; the setting standard of the lower threshold is: the balance score mean minus the balance score standard deviation; the calculation logic of the balance score mean is as follows: , where avg(B score ) is the balance score mean in the game within a period of time, is the balance score calculated for the i-th time; the calculation logic of the balance score standard deviation is as follows: where std(Bscore ) is the standard deviation of the balance score in the game over a period of time; specifically, avg(B score ) is used to reflect the overall balance of the game during a certain period; std(B score ) is used to reflect the fluctuation range of each balance score; the upper threshold is Up limitb = avg(B score ) + std(B score ), and the lower threshold is Low limitb = avg(B score ) - std(B score ). When the balance score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated; by calculating the mean and standard deviation of the balance score, it can help understand the current balance level and its fluctuation of the game.

[0079] The present invention is further configured to perform threshold judgment on the comprehensive abnormal behavior score. When the comprehensive abnormal behavior score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated; the setting standard of the upper threshold is: the mean of the comprehensive abnormal behavior score plus the standard deviation of the comprehensive abnormal behavior score; the setting standard of the lower threshold is: the mean of the comprehensive abnormal behavior score minus the standard deviation of the comprehensive abnormal behavior score; the calculation logic of the mean of the comprehensive abnormal behavior score is: wherein, avg(S anomaly ) is the mean of the comprehensive abnormal behavior score in the game over a period of time, is the comprehensive abnormal behavior score calculated for the i-th time; the calculation logic of the standard deviation of the comprehensive abnormal behavior score is: wherein, std(S anomaly 0 is the standard deviation of the comprehensive abnormal score in the game over a period of time; specifically, avg(S anomaly ) is used to measure the overall abnormal behavior level in the game system over a period of time; std(S anomaly ) is used to measure the degree of dispersion of the scores; the upper threshold is Up limits = avg(S anomaly ) + std(S anomaly ), and the lower threshold is Low limits = avg(S anomaly ) - std(S anomaly ). When the balance score exceeds the upper threshold or is lower than the lower threshold, an optimization signal is generated; the upper and lower thresholds are calculated based on the mean and standard deviation of the scores, providing a standard for automatically detecting and intervening in abnormal behaviors.

[0080] The present invention is further configured to obtain the above-mentioned balance score and comprehensive abnormal behavior score in real time through machine learning for visual display, and give an alarm when the balance score and comprehensive abnormal behavior score do not meet the threshold conditions; give an alarm when the mutation value is greater than the preset mutation value threshold, where the mutation value is the slope of the balance score and comprehensive abnormal behavior score.

[0081] Embodiment 2

[0082] Please refer to Figure 2 , the exemplary game test optimization system based on machine learning includes:

[0083] Data acquisition module: used to acquire test data, where the test data includes game balance data and informal gameplay data;

[0084] Preprocessing and construction module: used to preprocess the test data and construct confrontation balance features and deviant behavior features according to the preprocessed test data;

[0085] Calculation module: used to calculate the balance score according to the confrontation balance features and calculate the comprehensive abnormal behavior score according to the deviant behavior features;

[0086] Threshold judgment module: used to judge the threshold of the balance score, and generate a first optimization signal when the threshold condition is exceeded, and judge the threshold of the comprehensive abnormal behavior score, and generate a second optimization signal when the threshold condition is exceeded.

[0087] It should be noted that the game test optimization system based on machine learning provided in the above embodiment and the game test optimization method based on machine learning provided in the above embodiment belong to the same concept. The specific ways in which each module and unit perform operations have been described in detail in the method embodiment and will not be repeated here. In practical applications, the game test optimization system based on machine learning provided in the above embodiment can, according to needs, allocate the above functions to different functional modules, that is, divide the internal structure of the system into different functional modules to complete all or part of the functions described above, and this will not be limited here either.

[0088] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, or magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0089] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. Additionally, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context.

[0090] In this application, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0091] It should be understood that in various embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not imply the order of execution. The order of execution of each process should be determined based on its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0092] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0093] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0094] In several embodiments provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings, direct couplings, or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0095] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0096] In addition, the functional units in each embodiment of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0097] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0098] As described above, the above are only specific implementation manners of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for optimizing game testing based on machine learning, characterized in that, Including: Obtain test data, where the test data includes game balance data and informal gameplay data; Preprocess the test data, and construct confrontation balance features and deviant behavior features based on the preprocessed test data; Calculate a balance score based on the confrontation balance features, and calculate a comprehensive abnormal behavior score based on the deviant behavior features; Perform a threshold judgment on the balance score, and generate a first optimization signal when the threshold condition is exceeded. Perform a threshold judgment on the comprehensive abnormal behavior score, and generate a second optimization signal when the threshold condition is exceeded.

2. The machine learning-based game test optimization method according to claim 1, wherein The game balance data includes confrontation combinations used by players, player win-loss records, combination usage frequencies, combination battle data, and confrontation combination effects. The informal gameplay data includes player activity duration, player activity level, number of abnormal behaviors, and abnormal behavior frequencies; The confrontation balance features include combination usage frequency, combination relative strength, and combination effect difference. The deviant behavior features include informal behavior frequency and type, behavior abnormality degree, and system impact of abnormal behavior.

3. The method for optimizing game testing based on machine learning according to claim 2, wherein The combination usage frequency includes usage frequency and win rate difference; The combination relative strength includes relative strength and confrontation balance degree; The combination effect difference includes effect deviation and combination diversity; Calculate a balance score based on the usage frequency, the win rate difference, the relative strength, the confrontation balance degree, the effect deviation, and the combination diversity.

4. The machine learning-based game test optimization method according to claim 3, characterized in that The calculation logic of the usage frequency is as follows: Among them, F use is the usage frequency, is the usage frequency of the i-th combination of character, weapon and skill, is the player activity, and λ1 is the influence coefficient for adjusting the impact between the usage frequency and the player activity; The calculation logic of the winning rate difference is as follows: Among them, D win is the winning rate difference, is the winning rate of the i-th combination, W average is the average winning rate of all combinations, and λ2 is the influence coefficient between the winning rate and the usage frequency; The calculation logic of the relative strength is as follows: where R strength is the relative strength, is the number of times the i-th element participates in the confrontation, and γ1 is the influence coefficient between the adjustment strength and the confrontation; The calculation logic of the confrontation balance is as follows: Among them, B combat is the confrontation balance, is the number of confrontations of the i-th combination, is the failure rate during the confrontation of the i-th combination, and γ2 is the influence coefficient between the confrontation balance and the player activity; The calculation logic of the said effect deviation is as follows: Among them, E bias is the effect deviation, is the effect value of the i-th combination of character, weapon and skill, and E average is the average value of the effect values of all combinations, is the usage frequency of the i-th combination, and δ1 is the influence coefficient between the effect deviation and the usage frequency; The calculation logic of the combined diversity is as follows where T max is the maximum number of uses of all combinations in the system, and δ2 is the influence coefficient between the effect and the usage frequency; The calculation logic of the balance score is as follows: Among them, B score is the balance score, is the usage frequency of the i-th character, weapon or skill, is the winning rate difference of the i-th combination, is the relative strength of the i-th combination, is the confrontation balance degree of the i-th combination, is the effect deviation of the i-th combination, is the combination diversity of the i-th combination, and w1, w2, w3, w4, w5 and w6 are weight coefficients.

5. The method for optimizing game testing based on machine learning according to claim 2, wherein, The informal behavior frequency and type include abnormal behavior frequency and abnormal behavior type; The behavior abnormality degree includes behavior abnormality degree and behavior deviation degree; The system impact of the abnormal behavior includes system impact degree and vulnerability exploitation impact degree; Calculate a comprehensive abnormal behavior score based on the abnormal behavior frequency, the abnormal behavior type, the behavior abnormality degree, the behavior deviation degree, the system impact degree, and the vulnerability exploitation impact degree.

6. The machine learning-based game test optimization method according to claim 5, characterized in that The calculation logic of the abnormal behavior frequency is as follows: Among them, F anomaly is the abnormal behavior frequency, is the number of abnormal behaviors in the i-th time, is the playing time of the player, and α1 is the attenuation coefficient between the abnormal behavior and the session duration; The calculation logic for the abnormal behavior type is as follows: Among them, Tanomaly is the abnormal behavior type, βi is the weight coefficient of behavior type i, is the severity score of the i-th behavior, is the player's behavior pattern, and α2 is the relationship coefficient between the player's behavior pattern and the abnormal behavior type; The calculation logic of the behavior abnormality degree is as follows: where A anomaly is the behavior abnormality degree, is the complexity score of the i-th abnormal behavior, is the time point when the i-th abnormal behavior occurs, β1 is the influence coefficient between the abnormal behavior time point and the complexity, is the complexity score of the i-th normal behavior; The calculation logic of the behavior deviation degree is as follows: Among them, P dewi2tion is the behavior deviation degree, is the deviation degree between the i-th abnormal behavior and the normal behavior, is the complexity of the task when the i-th abnormal behavior occurs, is the measurement value of the i-th normal behavior; κ1 is the coefficient between the task complexity and the behavior deviation degree; The calculation logic of the system impact degree is as follows: Among them, I impact is the system impact degree, is the impact score of the i-th abnormal behavior on the game system, is the occurrence frequency of the i-th abnormal behavior, is the occurrence frequency of the i-th normal behavior; μ1 is the system impact degree coefficient. The calculation logic of the exploit impact is as follows: Among them, V exploit is the exploit impact, is the severity score of the i-th exploit, is the player activity of the i-th exploit, is the score of the i-th normal behavior, and μ2 is the coefficient between the exploit impact and the activity; The calculation logic of the comprehensive abnormal behavior score is as follows: Among them, S anomaly is the comprehensive abnormal behavior score, is the frequency of the i-th abnormal behavior, is the type and severity of the i-th abnormal behavior, is the abnormality degree of the i-th abnormal behavior, is the behavior deviation degree of the i-th time, is the impact degree of the i-th abnormal behavior on the system, is the impact degree of the i-th vulnerability exploitation, and ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 are weight coefficients.

7. The machine learning-based game test optimization method according to claim 1, wherein Perform a threshold judgment on the balance score. When the balance score exceeds the upper threshold or is lower than the lower threshold, generate an optimization signal; The setting standard of the upper threshold is: the mean value of the balance score plus the standard deviation of the balance score; The setting standard of the lower threshold is: the mean value of the balance score minus the standard deviation of the balance score; The calculation logic of the average balance score is as follows: where avg(B score ) is the average balance score in the game over a period of time, is the balance score calculated for the i-th time; The calculation logic of the standard deviation of the balance score is as follows: where std(B score ) is the standard deviation of the balance score in the game over a period of time.

8. The method for optimizing game testing based on machine learning according to claim 1, wherein Perform a threshold judgment on the comprehensive abnormal behavior score. When the comprehensive abnormal behavior score exceeds the upper threshold or is lower than the lower threshold, generate an optimization signal; The setting standard of the upper threshold is: the mean value of the comprehensive abnormal behavior score plus the standard deviation of the comprehensive abnormal behavior score; The setting standard of the lower threshold is: the mean value of the comprehensive abnormal behavior score minus the standard deviation of the comprehensive abnormal behavior score; The calculation logic of the mean value of the comprehensive abnormal behavior score is as follows: where avg(S anomaly ) is the mean value of the comprehensive abnormal behavior score in the game over a period of time, is the comprehensive abnormal behavior score calculated for the i-th time; The calculation logic of the standard deviation of the comprehensive abnormal behavior score is as follows: where std(S anomaly ) is the standard deviation of the comprehensive abnormal score in the game over a period of time.

9. The method for optimizing game testing based on machine learning according to claim 1, wherein Obtain the above balance score and comprehensive abnormal behavior score through machine learning in real time for visual display. When the balance score and the comprehensive abnormal behavior score do not meet the threshold conditions, give an early warning; when the mutation value is greater than the preset mutation value threshold, give an early warning, where the mutation value is the slope of the balance score and the comprehensive abnormal behavior score.

10. A game testing optimization system based on machine learning, which is used to implement the game testing optimization method based on machine learning according to any one of claims 1-9, characterized in that Including: Data acquisition module: used to obtain test data, where the test data includes game balance data and informal gameplay data; Preprocessing and construction module: used to preprocess the test data and construct adversarial balance features and deviant behavior features based on the preprocessed test data; Calculation module: used to calculate a balance score based on the adversarial balance features and calculate a comprehensive abnormal behavior score based on the deviant behavior features; Threshold judgment module: used to judge the balance score against a threshold, and generate a first optimization signal when the threshold condition is exceeded; used to judge the comprehensive abnormal behavior score against a threshold, and generate a second optimization signal when the threshold condition is exceeded.

Images