Game behavior analysis system based on deep learning

Abstract

The invention relates to the technical field of behavior analysis, in particular to a game behavior analysis system based on deep learning, which comprises a game activeness analysis module, a game social behavior analysis module, a user preference identification module and an update influence analysis module. According to the method, the daily active mode and the long-term participation trend of the user are comprehensively evaluated by utilizing the user game behavior data, the understanding of the user behavior is deeper and more accurate, the importance of each user can be evaluated by integrating the behavior data of the user and the social behavior result, the important game user can be identified, and the user experience is improved. The key game content is identified according to the importance of the user and the game behavior data, the key content preferred by the user can be updated during subsequent updating, the satisfaction and loyalty of the user are improved, the actual updating effect is judged after the game is updated, and a scientific basis is provided for future game design and optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of behavior analysis, and particularly to a game behavior analysis system based on deep learning. Background Art

[0002] The technical field of behavior analysis encompasses a wide range of methods and techniques for studying, evaluating, and predicting the behavior of individuals or groups. This field combines the theories and technologies of psychology, sociology, human-computer interaction, and computer science. By adopting machine learning, statistical analysis, pattern recognition, and algorithm optimization, behavior analysis techniques can extract meaningful behavior patterns and trends from complex data. The technology is widely applied in fields such as market research, user experience design, security monitoring, and health management, helping organizations and enterprises better understand customer or user behavior, thereby optimizing product design and service provision.

[0003] Among them, a game behavior analysis system is a system for monitoring and analyzing the behavior patterns of players in video games. The main purpose of the system is to understand players' behavior habits, preferences, and their changing trends by capturing and analyzing players' interaction data. This information can be used for iterative optimization of game design to improve players' gaming experience and satisfaction. At the same time, by deeply understanding players' behavior, developers can effectively adjust game mechanisms to enhance players' engagement and loyalty.

[0004] Traditional systems focus on the collection and simple analysis of static data, failing to deeply explore the behavior patterns and user needs behind the data, resulting in the inability to directly apply the analysis results to the actual improvement of game design, restricting the response speed and innovation ability of game development. Traditional systems ignore the subtle changes in user behavior, such as changes in communication frequency and dialogue content in games, lacking analysis tools for sensitivity to dynamic changes, making it impossible for developers to quickly identify and respond to changes in user behavior, leading to user loss and the loss of market opportunities. Traditional systems lack the analysis of user behavior before and after game updates, which is particularly prominent in the game industry that requires rapid feedback and adjustment, resulting in a lag in feedback on players' behavior, thus affecting players' overall gaming experience and satisfaction. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose a game behavior analysis system based on deep learning.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A game behavior analysis system based on deep learning, the system includes: The game activity analysis module, based on the game behavior data of users, counts the daily play time of users, analyzes the average weekly login times of users, evaluates the activity of users, analyzes the activity data of users, evaluates the stability of users, and obtains the game activity analysis results; The game social behavior analysis module, based on the conversation behavior data of users in the game, analyzes the conversation information of users in the game through a deep learning model, identifies the conversation types of users, evaluates the impact of communication behavior of users in the game, combines the number of friends of users in the game, evaluates the social impact of users in the game, and obtains the game social behavior analysis results; The user preference recognition module, based on the game activity analysis results and game social behavior analysis results of users, evaluates the importance level of each user according to the activity, stability and social impact of users, extracts the behavior data of users, evaluates the preference degree of the overall users for differentiated game content, identifies the game content preferred by users, and obtains the user preference content recognition results; The update impact analysis module, based on the user preference content recognition results, extracts the changes in user activity and stability before and after the game update, combines the changes in the play frequency and time of users for differentiated game content, evaluates the game update effect, and obtains the game update impact analysis results.

[0007] The improvement of the present invention is that the step of evaluating the activity of users is as follows: Based on the game behavior data of users, count the play time of users, and analyze the average weekly login times of users, to obtain activity-related data; Based on the activity-related data, through the formula: ; Calculate the activity index of users; Wherein, is the daily play time of users, is the total number of login times of users in the most recent week, and are weight coefficients, is the activity index of users; Based on the activity index of users, according to the size of the activity index, evaluate the activity of users to obtain user activity information.

[0008] The improvement of the present invention is that the step of obtaining the game activity analysis results is as follows: Based on the user activity information, at preset time intervals, extract the activity data of users over a period of time to obtain stability-related information; Based on the stability-related information, through the formula: ; Calculate the stability index of the user; Among them, is the stability index of the user, is the total number of days within the statistical period, is the user activity index on the day, is the average activity index;

[0009] An improvement of the present invention is that the step of evaluating the influence of the user's communication behavior in the game is: Based on the dialogue behavior data of the user in the game, through a deep learning model, extract the keywords in the user's dialogue information, analyze the dialogue information of the user in the game, identify the types of the user's conversations, including normal game communication, normal life conversations, and abnormal conversations, and obtain the user conversation type recognition result; Based on the user conversation type recognition result, count the frequency of each type of the user's conversation to obtain communication influence correlation data; Based on the communication influence correlation data, through the formula: ; Calculate the communication influence index of the user to obtain the user communication influence evaluation result; Among them, is the normal game communication frequency of the user, is the normal life conversation frequency of the user, is the abnormal conversation frequency of the user, , and are influence factors, is the communication influence index of the user.

[0010] An improvement of the present invention is that the step of obtaining the user social behavior analysis result is: Based on the user communication influence evaluation result, extract the number of the user's game friends, and calculate the average number of friends of all users to obtain social influence correlation data; Based on the social influence correlation data, through the formula: ; Calculate the social influence index of the user to obtain the user social behavior analysis result; Among them, is the average number of friends of the game users, is the communication influence index of the user, is the number of friends of the user, is the average activity of the user's friends, is the social influence index of the user.

[0011] The improvement of the present invention is that the step of evaluating the importance level of each user is: Based on the user activity analysis result and the user social behavior analysis result, extract the activity index, stability index and social influence index of the user, and extract the preset minimum importance level to obtain importance correlation data; Based on the importance correlation data, through the formula: ; Calculate the importance level of the user to obtain the user importance evaluation result; Wherein, is the preset minimum importance level, is the activity index of the user, is the stability index of the user, is the social influence index of the user, , and are weight coefficients, is the importance level of the user.

[0012] The improvement of the present invention is that the step of obtaining the user preference content recognition result is: Based on the user importance evaluation result, by counting the number of users with different importance levels, analyzing the proportion of users with different importance levels, and extracting the behavior data of users with different importance levels, including the frequency and time of playing each game content, obtain game preference correlation data; Based on the game preference correlation data, through the formula: ; Calculate the preference index of the user for the game content; Wherein, is the proportion of users with the th importance level, is the number of participations of users with the th importance level in the game content, is the total playing time of users with the th importance level in the game content, is the value of the th importance level, is the total number of importance levels, is the preference index of the user for the game content; Based on the user's preference index for game content, sort the game content according to the magnitude of the preference index, identify the game content preferred by the user, and obtain the recognition result of the user's preferred content.

[0013] The improvement of the present invention is that the step of obtaining the game update impact analysis result is as follows: Based on the recognition result of the user's preferred content, implement game updates according to the user's preferred content, extract the user activity index and stability index data before and after the update, and extract the changes in the user's participation in the target game content, including changes in play frequency and time, to obtain update effect correlation data; Based on the update effect correlation data, through the formula: ; Calculate the game update effect score; Wherein, is the game update effect score, is the change in the activity index, is the change in the stability index, is the change in the game participation frequency of the target update content, is the change in the game time of the target update content, 、 and are weight coefficients, is the sign function, returning the sign of, when is a positive number, is positive, when is a negative number, is negative; Based on the game update effect score, evaluate the game update effect according to the positive / negative and magnitude of the score, and judge whether the game update effect meets the expectation to obtain the game update impact analysis result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by utilizing user game behavior data, the daily active patterns and long-term participation trends of users are comprehensively evaluated, and the understanding of user behavior is deeper and more accurate. Through in-depth analysis of in-game dialogue content, the communication patterns of users and their social influence are identified. By integrating user behavior data and social behavior results, the importance of each user can be evaluated, important game users can be identified, and based on the importance of users and game behavior data, key game content can be identified. Then, during subsequent updates, updates can be made to the key content preferred by users, improving user satisfaction and loyalty. After the game is updated, through the pre- and post-comparative analysis of activity and stability, as well as the evaluation of changes in engagement, the actual effect of the update can be accurately judged, providing a scientific basis for future game design and optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is the system flowchart of the present invention; Figure 2 is the flowchart for evaluating the activity of users in the present invention; Figure 3 is the flowchart for obtaining the analysis results of user activity in the present invention; Figure 4 is the flowchart for evaluating the influence of communication behavior of users in the game in the present invention; Figure 5 is the flowchart for the analysis results of user social behavior in the present invention; Figure 6 is the flowchart for evaluating the importance level of each user in the present invention; Figure 7 is the flowchart for obtaining the recognition results of user-preferred content in the present invention; Figure 8 is the flowchart for obtaining the analysis results of the impact of game updates in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0017] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more, unless otherwise specifically defined.

[0018] Please refer to Figure 1 , the present invention provides a technical solution: a game behavior analysis system based on deep learning, the system includes: The game activity analysis module is based on the game behavior data of users, counts the daily play time of users, analyzes the average weekly online times of users, evaluates the activity of users, and analyzes the activity data of users to evaluate the stability of users, so as to obtain the user activity analysis result; The game social behavior analysis module is based on the dialogue behavior data of users in the game. Through a deep learning model, it extracts the keywords in the user dialogue information, analyzes the dialogue information of users in the game, identifies the dialogue types of users, including normal game communication, normal life dialogue, and abnormal dialogue. Combining the dialogue frequency of users in the game, it evaluates the influence of the communication behavior of users in the game. Combining the number of friends of users in the game, it evaluates the social influence of users in the game, so as to obtain the user social behavior analysis result; The user preference recognition module is based on the user activity analysis result and the user social behavior analysis result. According to the activity, stability, and social influence of users, it evaluates the importance level of each user, and extracts the behavior data of users, including the frequency and time of playing different game contents. Combining the proportion of users with different importance levels, it evaluates the preference degree of the overall users for different game contents, identifies the game contents preferred by users, so as to obtain the user preference content recognition result; The update impact analysis module is based on the user preference content recognition result, extracts the changes in user activity and stability before and after the game update, combines the changes in the playing frequency and time of users for different game contents, evaluates the game update effect, and obtains the game update impact analysis result.

[0019] The results of user activity analysis include the average daily playtime, the average weekly login times, the activity score, and the stability score. The results of user social behavior analysis include the classification information of conversation types, the conversation frequency, the social network density, and the social influence score. The results of identifying user preferred content include game content preferences, playtime information, and frequency ranking information. The analysis results of game update impacts include the activity change rate, the stability change index, the user retention rate, and the comparison of playtimes before and after the update.

[0020] Please refer to Figure 2 , and the steps to evaluate the user's activity are as follows: Based on the user's game behavior data, count the user's playtime and analyze the user's average weekly login times to obtain activity-related data; Based on the activity-related data, through the formula: ; Calculate the user's activity index, is the user's activity index; where, is the user's daily playtime, is the total number of login times of the user in the most recent week, and are the weight coefficients; Based on the user's activity index, evaluate the user's activity according to the size of the activity index to obtain the user activity information.

[0021] Formula: ; Meaning and acquisition method of parameters: : The user's daily playtime, which is obtained by recording the total daily playtime of the user.

[0022] : The total number of login times of the user in the most recent week. This is obtained by recording the daily login times of a user in a week and accumulating them. For example, if a user's login times in a week are 7, 9, 8, 10, 9, 11, 12 times, then will be the total number of login times in this week, that is, 7 + 9 + 8 + 10 + 9 + 11 + 12 = 66 times.

[0023] and : The weight coefficients, which are used to adjust the contributions of the daily playtime and the weekly login times in the calculation of the activity index. The selection of the weight coefficients depends on the regression analysis of historical data, and usually the best values are determined by minimizing the prediction error.

[0024] Calculation example: Set the data of a user: The user's daily playing time is 15 hours per day, the total number of logins of this user in the recent week is 66 times, and the weight coefficient determined through historical data analysis is and .

[0025] Activity index The calculation formula is: ; Calculate the logarithmic part of the daily playing time: ; Calculate the square root of the number of logins in a week: ; Substitute the above results into the formula: ; The activity index represents the activity level of this user. The higher this value, the more active the user is. This value is obtained through weighted calculation by combining the user's playing time and login frequency, and can more comprehensively reflect the user's activity.

[0026] Please refer to Figure 3 , and the steps to obtain the user activity analysis result are: Based on the user activity information, extract the user activity data within a period of time at a preset time interval to obtain the stability correlation information; Based on the stability correlation information, through the formula: ; Calculate the stability index of the user; Among them, is the stability index of the user, is the total number of days within the statistical period, is the user activity index on the day, is the average activity index;

[0027] Formula: ; Detailed explanation of parameters and acquisition method: is the total number of days within the statistical period, such as the number of days in a month or six months, usually determined according to the available range of user activity data.

[0028] is the activity index of the user on the day, obtained through the calculation method in the previous steps.

[0029] is the average activity index, expressed as , which provides the average level of user activity and is used to measure the deviation of an individual from the average level.

[0030] Calculation example: Set the activity indices of a user over four days to be , , , .

[0031] Calculate the average activity : ; Calculate the square of the difference between the daily activity index and the average value each day, and then sum them up:

[0032]

[0033] ; Calculate : ; The calculation result indicates that the user's activity has low volatility, thus indicating that the user's behavior is relatively stable during the inspection period. A high stability index indicates that the user's daily activity pattern is relatively consistent and has low volatility.

[0034] Please refer to Figure 4 , the steps to evaluate the impact of a user's communication behavior in the game are as follows: Based on the user's dialogue behavior data in the game, through a deep learning model, extract the keywords in the user's dialogue information, analyze the user's dialogue information in the game, identify the user's dialogue types, including normal game communication, normal life dialogue, and abnormal dialogue, and obtain the user dialogue type recognition result; Based on the user dialogue type recognition result, count the frequency of each user dialogue type to obtain the communication impact correlation data; Based on the communication impact correlation data, through the formula: ; Calculate the user's communication impact index to obtain the user communication impact evaluation result; Among them, is the frequency of the user's normal game communication, is the frequency of the user's normal life dialogue, is the frequency of the user's abnormal dialogue, , and is the impact factor, and is the user's communication impact index.

[0035] Formula: ; Detailed Explanation of Parameters and Acquisition Method is the normal game communication frequency of the user, is the normal life conversation frequency of the user, and is the abnormal conversation frequency of the user. The frequency is calculated by dividing the number of entries of each type of conversation by the corresponding time period (such as the total time within a day). The calculation method of can reflect the activity level of the user in different types of conversations.

[0036] 、 、 : The impact factors corresponding to each type of conversation, which reflect the weights of different types of conversations in the evaluation of the user's communication behavior. The factors are preset according to the social impact and importance of the conversation type, usually determined based on historical data or expert opinions.

[0037] Calculation Example: Set the following data: within a day, the number of normal game conversation entries of a user is 100, the number of normal life conversation entries is 150, and the number of abnormal conversation entries is 10. The time period is 24 hours. The conversation frequencies are calculated as follows: ; ; ; Set the impact factors to be , , 。

[0038] Calculate :

[0039]

[0040]

[0041]

[0042] ; The calculation result represents the overall communication impact index of the user, which reflects the quality and magnitude of the user's social interaction within the game. A higher value indicates that the user has a stronger communication influence in the game.

[0043] Please refer to Figure 5 for the steps to obtain the analysis results of the user's social behavior: Based on the user communication impact assessment results, extract the number of the user's in-game friends, calculate the average number of friends of all users, and obtain the social impact correlation data; Based on the social impact correlation data, through the formula: ; Calculate the user's social impact index and obtain the user social behavior analysis results; Among them, is the average number of friends of in-game users, is the user's communication impact index, is the number of the user's friends, is the average activity of the user's friends, is the user's social impact index.

[0044] Formula: ; Detailed parameter explanations and acquisition methods: is the user's communication impact index, obtained through the previous steps.

[0045] is the number of the user's friends, directly extracted by accessing the in-game social network database, and the total number of the user's current friends is counted.

[0046] is the average activity of the user's friends, obtained by aggregating the activities of all the user's friends and then taking the average.

[0047] is the average number of friends of in-game users, obtained by querying the number of friends of all users in the game in the database and then calculating the average value for standardized comparison.

[0048] Calculation example: Set the user's communication impact index , the number of the user's friends , the average number of friends in the whole game , the average activity of friends .

[0049] Calculate the friend number ratio: ; Calculate the logarithmic impact of friend activity: 8; Calculate : ; The calculation result represents the user's social influence index, which quantifies the user's social influence in the game. A higher value indicates that the user not only communicates frequently in the game but also has many active friends, thus having a greater influence in the game's social network. This method provides a comprehensive assessment of the user's in-game social behavior by combining the quality of communication, social breadth, and social depth of the user.

[0050] Please refer to Figure 6 , the steps to evaluate the importance level of each user are as follows: Based on the user activity analysis results and user social behavior analysis results, extract the user's activity index, stability index, and social influence index, and extract the preset minimum importance level to obtain importance correlation data; Based on the importance correlation data, through the formula: ; Calculate the importance level of the user to obtain the user importance evaluation result; Among them, is the preset minimum importance level, is the user's activity index, is the user's stability index, is the user's social influence index, , and are weight coefficients, is the importance level of the user.

[0051] Formula: ; Parameter details and acquisition methods: : The user's activity index, obtained by calculation in the previous steps.

[0052] : The user's stability index, obtained by calculation in the previous steps.

[0053] : The user's social influence index, obtained by calculation in the previous steps.

[0054] , , : The weight coefficients represent the relative importance of each index in calculating the overall influence, usually set according to business requirements and historical data.

[0055] : The preset minimum importance level, ensuring that the user level is not lower than this value to fairly reflect the basic influence of all users.

[0056] Calculation example: Set the activity index of a user , stability index , social influence index , and set the weight coefficients to be , , , minimum importance level .

[0057] The calculation process is as follows: Calculate the weighted sum of each index:

[0058]

[0059]

[0060] ; Convert the value to logarithmic form to increase the non-linear response, and calculate the integer part: ; ; Minimum importance level limit: ; The calculation results show that although the user shows a certain degree of activity and social influence in the game, considering the preset minimum level, their final importance level is set to 2. This ensures that all users have at least basic recognition and importance, avoiding low ratings caused by algorithm sensitivity. This calculation method aims to balance the comprehensive evaluation of user contributions and maintain a certain level baseline.

[0061] Please refer to Figure 7 , the steps to obtain the recognition result of user-preferred content are as follows: Based on the user importance evaluation result, by counting the number of users with different importance levels, analyzing the proportion of users with different importance levels, and extracting the behavioral data of users with different importance levels, including the frequency and time of playing each game content, obtain the game preference correlation data; Based on the game preference correlation data, through the formula: ; Calculate the preference index of the user for the game content; Among them, is the proportion of users at the th importance level, is the The number of times users of a certain importance level participate in game content, is the total playing time of users of the th importance level in game content, is the value of the th importance level, is the total number of importance levels, is the user's preference index for game content;

[0062] Formula: ; Detailed parameter explanations and acquisition methods: : The proportion of users of the th importance level. The proportion is obtained by analyzing the user database and calculating the proportion of the number of users in each importance level to the total number of users.

[0063] : The number of times users of the th importance level participate in game content, that is, the number of times users participate in specific game content. This is usually counted by tracking the user's game behavior logs.

[0064] : The total playing time of users of the th importance level in game content, that is, the total playing time of users in specific game content. It is also counted by analyzing the user's game behavior logs.

[0065] : The value of the th importance level, obtained by the method calculated in the previous steps.

[0066] : The total number of importance levels, which is determined according to the user classification strategy of the game or platform.

[0067] Calculation example: Suppose there are three different importance levels ( ), and the proportion of users, participation times, playing time, and level values at each level are as follows: Level 1: , hours, ; Level 2: , hours, 2; Level 3: , hours .

[0068] The calculation process is as follows:

[0069]

[0070]

[0071]

[0072] ; The calculation results show that the game content is very popular among users of different levels, especially among user groups with a higher level of importance. This calculation method provides an in-depth understanding of the preference levels of the game content among different user levels, helping game developers and marketing strategists optimize products and services for different user groups.

[0073] Please refer to Figure 8 for the steps to obtain the analysis results of the impact of game updates: Based on the recognition results of user-preferred content, implement game updates according to the user's preferred content, extract the user activity index and stability index data before and after the update, and extract the changes in the user's participation in the target game content, including changes in play frequency and time, to obtain update effect correlation data; Based on the update effect correlation data, through the formula: ; Calculate the game update effect score; Among them, is the game update effect score, is the change in the activity index, is the change in the stability index, is the change in the game participation frequency of the target update content, is the change in the game time of the target update content, , and are weight coefficients, is the sign function, which returns the sign of, when is a positive number, is positive, when is a negative number, is negative; Based on the game update effect score, evaluate the game update effect according to the positive / negative and magnitude of the score, and judge whether the game update effect meets the expectations to obtain the analysis results of the impact of game updates.

[0074] Formula: ; Detailed Explanation and Acquisition Method of Parameters: : The change in the activity index, obtained by comparing the average activity indices of users before and after the game update, calculated as the index after the update minus the index before the update.

[0075] : The change in the stability index. Similar to the activity, this is obtained by comparing the average stability indices of users before and after the update.

[0076] and : is the change in the game participation frequency of the target updated content, is the change in the game time of the target updated content. These changes are obtained through log analysis by statistically analyzing the differences in the average play time and average optimization frequency of all users on specific game content before and after the update.

[0077] : The sign function, used to determine the positive or negative sign of. This helps to correctly handle the impact of increases and decreases in frequency and time on the total score.

[0078] 、 、 : Weight coefficients, corresponding to the influence of activity change, stability change, and game participation change respectively. The coefficients are obtained based on business requirements and historical data analysis to ensure the appropriate influence of each factor in the scoring.

[0079] Calculation Example: Set the indices of users before and after the game update as follows: The change in the activity index before and after the update , The change in the stability index before and after the update , The change in the participation frequency of the target game content , The change in the play time of the target game content hours. Set the weight coefficients as: , , .

[0080] Calculate the game update effect score :

[0081]

[0082]

[0083]

[0084] ; Calculation result , indicating that the game update has had a negative impact, especially a significant decrease in the frequency and time of game participation. This negative result reflects that the game update may not have achieved the expected effect of attracting users, or the updated content may not meet the preferences of this user.

[0085] The above are only the preferred embodiments of the present invention, and the present invention is not limited to other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A game behavior analysis system based on deep learning, characterized in that, The system includes: The game activity analysis module, based on the user's game behavior data, counts the user's daily play time, analyzes the user's average weekly log-in times, evaluates the user's activity, and analyzes the user's activity data to evaluate the user's stability, obtaining the user activity analysis result; The game social behavior analysis module, based on the user's conversation behavior data in the game, analyzes the user's conversation information in the game through a deep learning model, identifies the user's conversation type, evaluates the impact of the user's communication behavior in the game, and combines the number of friends of the user in the game to evaluate the social impact of the user in the game, obtaining the user social behavior analysis result; The user preference recognition module, based on the user activity analysis result and the user social behavior analysis result, evaluates the importance level of each user according to the user's activity, stability, and social impact, extracts the user's behavior data, evaluates the preference degree of the overall users for differentiated game content, and identifies the game content preferred by the users, obtaining the user preference content recognition result; The update impact analysis module, based on the user preference content recognition result, extracts the changes in user activity and stability before and after the game update, combines the changes in the play frequency and time of the user for differentiated game content, evaluates the game update effect, obtaining the game update impact analysis result.

2. The game behavior analysis system based on deep learning according to claim 1, characterized in that The steps for evaluating the user's activity are as follows: Based on the user's game behavior data, count the user's play time and analyze the user's average weekly log-in times to obtain activity-related data; Based on the activity-related data, through the formula: ； Calculate the user's activity index; wherein, is the user's daily play time, is the total number of logins of the user within the most recent week, and is the weight coefficient, is the user's activity index; Based on the user's activity index, evaluate the user's activity according to the size of the activity index to obtain the user activity information.

3. The game behavior analysis system based on deep learning according to claim 2, characterized in that, The steps for obtaining the user activity analysis result are as follows: Based on the user activity information, extract the user's activity data within a period of time at a preset time interval to obtain stability-related information; Based on the stability-related information, through the formula: ； Calculate the user's stability index; Among them, is the stability index of the user, is the total number of days within the statistical period, is the user activity index on the day, and is the average activity index; Based on the user's stability index, evaluate the user's stability according to the size of the stability index, and integrate the user's activity index to obtain the user activity analysis result.

4. The game behavior analysis system based on deep learning according to claim 1, characterized in that, The steps for evaluating the impact of the user's communication behavior in the game are as follows: Based on the user's conversation behavior data in the game, through a deep learning model, extract the keywords in the user conversation information, analyze the user's conversation information in the game, and identify the user's conversation type, including normal game communication, normal life conversation, and abnormal conversation, obtaining the user conversation type recognition result; Based on the user conversation type recognition result, count the frequency of each user conversation type to obtain communication impact-related data; Based on the communication impact-related data, through the formula: ； Calculate the user's communication impact index to obtain the user communication impact evaluation result; Among them, is the normal game communication frequency of the user, is the normal life conversation frequency of the user, is the abnormal conversation frequency of the user, , and are influencing factors, is the communication influence index of the user.

5. The game behavior analysis system based on deep learning according to claim 4, characterized in that The steps for obtaining the user social behavior analysis result are as follows: Based on the user communication impact evaluation result, extract the number of game friends of the user and calculate the average number of friends of all users to obtain social impact-related data; Based on the social impact-related data, through the formula: ； Calculate the social influence index of the user to obtain the user social behavior analysis result; Among them, is the average number of friends of game users, is the communication influence index of users, is the number of friends of users, is the average activity of users' friends, is the social influence index of users.

6. The game behavior analysis system based on deep learning according to claim 1, wherein, The step of evaluating the importance level of each user is as follows: Based on the user activity analysis result and the user social behavior analysis result, extract the user activity index, stability index, and social influence index of the user, and extract the preset minimum importance level to obtain importance correlation data; Based on the importance correlation data, through the formula: ； Calculate the importance level of the user to obtain the user importance evaluation result; wherein, is a preset minimum importance level, is the user's activity index, is the user's stability index, is the user's social influence index, , and are weight coefficients, is the user's importance level.

7. The game behavior analysis system based on deep learning according to claim 6, wherein The step of obtaining the user preferred content recognition result is as follows: Based on the user importance evaluation result, by counting the number of users with different importance levels, analyzing the proportion of users with different importance levels, and extracting the behavior data of users with different importance levels, including the frequency and time of playing each game content, obtain game preference correlation data; Based on the game preference correlation data, through the formula: ； Calculate the preference index of the user for the game content; Among them, is the proportion of users at the th importance level, is the number of times users at the th importance level participate in the game content, is the total playing time of users at the th importance level on the game content, is the value of the th importance level, is the total number of importance levels, is the preference index of users for the game content; Based on the preference index of the user for the game content, sort the game content according to the size of the preference index, identify the game content preferred by the user, and obtain the user preferred content recognition result.

8. The game behavior analysis system based on deep learning according to claim 1, characterized in that The step of obtaining the game update impact analysis result is as follows: Based on the user preferred content recognition result, implement game updates according to the user's preferred content, extract the user activity index and stability index data before and after the update, and extract the change in the user's participation in the target game content, including the change in play frequency and time, to obtain update effect correlation data; Based on the update effect correlation data, through the formula: ； Calculate the game update effect score; Among them, is the game update effect score, is the change in the activity index, is the change in the stability index, is the change in the game participation frequency of the target update content, is the change in the game time of the target update content, 、 and are weight coefficients, is the sign function, returning the sign of, when is a positive number, is positive, when is a negative number, is negative; Based on the game update effect score, evaluate the game update effect according to the positive / negative and size of the score, and determine whether the game update effect meets the expectation to obtain the game update impact analysis result.

Images