Game platform management method and system based on artificial intelligence

Abstract

The invention provides a game platform management method and system based on artificial intelligence. The method belongs to the technical field of game platform management, and comprises the following steps: acquiring multi-source data through multiple channels, constructing a user portrait based on the acquired multi-source data, training the user data through a deep learning model, and identifying user preferences, habits and potential demands; the server load, the game popularity and the user distribution are monitored in real time, the game peak period is pre-judged through time sequence analysis based on historical data, and resource allocation is dynamically adjusted according to a prediction result; through collection and analysis of multi-source data, an accurate user portrait is constructed, and in combination with deep learning and an emotion analysis model, preferences, emotion changes and potential demands of the user can be identified and understood, so that personalized game recommendation, customer service and content pushing are provided.

Description

Technical Field

[0001] The present invention proposes a game platform management method and system based on artificial intelligence, belonging to the technical field of game platform management. Background Art

[0002] Traditional game platform management methods mainly rely on manual monitoring and preset rules, which are difficult to cope with complex and changeable user behaviors, game content updates and security threats. With the rapid development of the game industry, the amount of data faced by game platforms has increased dramatically, and traditional methods have obvious deficiencies in processing efficiency, accuracy and response speed. Therefore, it is particularly important to develop a game platform management method that can automatically adapt and make intelligent decisions. Summary of the invention

[0003] The present invention provides a game platform management method and system based on artificial intelligence to solve the problems mentioned in the above background technology: The present invention proposes a game platform management method based on artificial intelligence, the method comprising: S1. Obtain multi-source data through multiple channels, build user portraits based on the acquired multi-source data, and train user data through deep learning models to identify user preferences, habits and potential needs; S2. Monitor server load, game popularity and user distribution in real time. Based on historical data, use time series analysis to predict game peak hours and dynamically adjust resource allocation according to the prediction results. S3. Use natural language processing and image recognition technology to conduct real-time review of game content, identify and filter bad information, and use AI algorithms to detect abnormal behavior and take defensive measures based on the detection results; S4. Based on natural language processing and GPT models, provide 24 / 7 intelligent customer service and adjust the response strategy according to user emotions.

[0004] The present invention proposes a game platform management system based on artificial intelligence, the system comprising: Data acquisition module: obtains multi-source data through multiple channels, builds user portraits based on the obtained multi-source data, and trains user data through deep learning models to identify user preferences, habits and potential needs; Real-time monitoring module: monitors server load, game popularity and user distribution in real time. Based on historical data, it uses time series analysis to predict game peak hours and dynamically adjusts resource allocation according to the prediction results. Anomaly detection module: Through natural language processing and image recognition technology, it conducts real-time review of game content, identifies and filters bad information, and uses AI algorithms to detect abnormal behaviors and take defensive measures based on the detection results; Strategy adjustment module: Based on natural language processing and GPT model, it provides 24 / 7 intelligent customer service and adjusts the response strategy according to user emotions.

[0005] Beneficial effects of the present invention: by collecting and analyzing multi-source data, building accurate user portraits, combining deep learning and sentiment analysis models, it is possible to identify and understand user preferences, emotional changes and potential needs, thereby providing personalized game recommendations, customer service and content push, which can improve user satisfaction, and effectively improve user stickiness and platform activity; by real-time monitoring and predicting game peak hours, combining ARIMA and other time series models to analyze historical data, it is possible to predict server load requirements in advance, dynamically adjust resource allocation, ensure that the game is kept smooth under high load conditions, reduce the risk of server downtime, optimize resource utilization efficiency, and reduce operating costs; with the help of natural language processing and image recognition technology, the system can identify and filter bad information, sensitive words and images in real time, and improve the content review efficiency of the platform. At the same time, through the real-time response of abnormal behavior detection and defense measures, the occurrence of cheating, fraud and other behaviors is reduced, and the fairness and security of the platform are guaranteed; combined with the GPT model and sentiment analysis technology, the platform can provide users with 24 / 7 intelligent customer service, monitor user emotions in real time and make corresponding adjustments, and improve the quality and satisfaction of user communication. Through multi-round dialogue technology and personalized emotional dialogue services, not only can instant answers be provided, but also adjustments can be made according to the emotional needs of users, thereby improving customer service response efficiency and service quality; through machine learning and data analysis, in-depth analysis of resource usage, user behavior and platform operations can be achieved, potential bottlenecks can be identified, resource scheduling strategies can be optimized, and management strategies can be continuously improved based on user feedback. This data-driven decision-making method enables the platform to remain flexible and competitive in a rapidly changing market environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a step diagram of the method of the present invention; Figure 2 This is a system module diagram of the present invention. DETAILED DESCRIPTION

[0007] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0008] One embodiment of the present invention, as Figure 1 As shown, a game platform management method based on artificial intelligence, the method comprising: S1. Obtain multi-source data through multiple channels, build user portraits based on the obtained multi-source data, and train user data through deep learning models to identify user preferences, habits and potential needs; combine natural language processing technology to conduct sentiment analysis on user comments and chat records, identify user emotional changes, and based on the analysis results, provide accurate basis for personalized recommendations and anti-addiction strategies, such as pushing games according to user preferences and adjusting push strategies according to emotional changes; S2. Monitor server load, game popularity and user distribution in real time. Based on historical data, use time series analysis to predict game peak hours. According to the prediction results, dynamically adjust resource allocation, such as increasing server capacity and optimizing network bandwidth to ensure smooth game operation. S3. Through natural language processing and image recognition technology, we can conduct real-time review of game content, identify and filter bad information, and use AI algorithms to detect abnormal behaviors, such as cheating and malicious attacks, and take defensive measures based on the detection results; S4. Based on natural language processing and GPT models, provide 24 / 7 intelligent customer service, adjust the response strategy according to user emotions, such as providing soothing responses to anger, identify and guide positive topics based on data analysis algorithms, such as pushing community activities based on user interests, promote community interaction, and continuously optimize customer service strategies and community management strategies based on user feedback.

[0009] The working principle of the above technical solution is as follows: First, the platform obtains multi-source data through multiple channels (such as user registration information, game behavior records, social media links, etc.); through the obtained multi-source data, a detailed user portrait is constructed, including the user's age, gender, game preferences, consumption habits, etc.; through the deep learning model, the user data is trained to identify the user's preferences, habits and potential needs. For example, by analyzing the user's historical game records, the user's favorite game type can be inferred; combined with natural language processing technology, sentiment analysis is performed on user comments and chat records to identify user emotional changes, such as positive, negative or neutral; based on the above analysis results, personalized game recommendations are provided to users, such as pushing new games according to user preferences. At the same time, the push strategy is adjusted according to the user's emotional changes to avoid pushing content that may cause discomfort when the user is in a low mood. In addition, the platform can also use these analyses to formulate anti-addiction strategies, such as sending reminders when the user's game time is too long; the platform monitors the server load, game popularity and user distribution in real time to obtain the latest game operation status; based on historical data, time series analysis technology is used to predict the peak hours of the game. For example, by analyzing the user login data of the past few months, it is possible to predict which time periods are the most frequent for users to log in; based on the prediction results, the platform dynamically adjusts resource allocation. Before the peak period arrives, the server capacity is increased, the network bandwidth is optimized, etc., to ensure that the game can run smoothly and avoid problems such as freezing or disconnection caused by insufficient resources; through natural language processing and image recognition technology, the platform conducts real-time review of game content, including identifying and filtering bad information, such as violence, pornography, etc.; using AI algorithms to detect abnormal user behavior, such as cheating, malicious attacks, etc. The user's behavior pattern is analyzed through AI algorithms to identify abnormal behaviors that are inconsistent with normal behavior; based on the detection results, the platform takes corresponding defensive measures. For example, for cheating behavior, the cheating account can be banned; for malicious attacks, security measures can be taken to protect the safety of other users; based on natural language processing and GPT models, the platform provides 24 / 7 intelligent customer service, and the intelligent customer service can adjust the response strategy according to the user's emotions, such as providing soothing responses to anger. Using data analysis algorithms, intelligent customer service can also identify and guide positive topics. For example, relevant community activities are pushed based on user interests to promote interaction and communication between users; based on user feedback and interaction data, the platform continuously optimizes customer service strategies and community management strategies, including improving the response quality of intelligent customer service and increasing the attractiveness of community activities.

[0010] The effects of the above technical solutions are as follows: through the training of user data by deep learning models, it is possible to accurately identify user preferences, habits and potential needs, thereby providing users with personalized game recommendations, increasing the possibility of users finding games of interest, and improving user satisfaction and loyalty; combining natural language processing technology for sentiment analysis, it is possible to capture user emotional changes in real time, and adjust the push strategy according to the emotion, avoiding pushing content that may cause discomfort when the user is in a low mood, while increasing the chance of pushing relevant content when the user is in a positive mood, further improving the user experience; through real-time monitoring of server load, game popularity and user distribution, and using time series analysis to predict the peak hours of the game, the platform can make resource adjustment decisions in advance, ensuring that the game can run smoothly during peak hours, avoiding problems such as freezes or disconnections due to insufficient resources, and improving operational efficiency; dynamic adjustment of resource allocation also means that resource investment can be reasonably reduced during low periods, thereby optimizing resources. Utilize and reduce costs; through natural language processing and image recognition technology, the platform can conduct real-time review of game content, effectively identify and filter out bad information, help maintain a healthy and positive gaming environment, and protect users from bad information; use AI algorithms to detect abnormal behaviors such as cheating and malicious attacks, and take defensive measures based on the detection results, which can effectively curb the occurrence of bad behaviors and ensure the fairness and security of the game; intelligent customer service based on natural language processing and GPT models can provide 24-hour customer service support and adjust response strategies according to user emotions, which improves customer service efficiency and enhances user satisfaction and trust; identify and guide positive topics through data analysis algorithms, such as pushing community activities based on user interests, which can promote interaction and communication between users and enhance the cohesion and activity of the community; continuously optimize customer service strategies and community management strategies based on user feedback to ensure that the platform always stays at the forefront of user needs and continuously improves user experience.

[0011] In one embodiment of the present invention, the S1 includes: S11, acquiring multi-source data from multiple channels, wherein the multiple channels include platform logs, user behavior records, and third-party data sources, and preprocessing the acquired multi-source data; S12. Build a preliminary user profile based on basic user information (such as age, gender, region), game preferences and consumption habits; and segment users based on clustering algorithms (such as K-means) to identify different user groups; S13. Combine user behavior time series analysis to build user behavior patterns, such as game duration and game type preferences, and train user data through deep learning models (such as LSTM and Transformer); S14. Through feature engineering, extract key features of user behavior as model input and perform cross-validation. Through natural language processing technology, pre-process user comments and chat records, including word segmentation and part-of-speech tagging; S15. Use sentiment lexicon or sentiment classification model (such as BERT-based model) to perform sentiment analysis on user text and monitor user emotion changes in real time, such as anger, disappointment, and excitement. S16. Analyze the reasons for emotional changes based on user portraits, such as game difficulty and social experience. Recommend games that meet users' preferences based on user portraits and deep learning model prediction results. S17. Adjust the push strategy according to changes in user emotions, such as avoiding pushing stimulating content when the user is in a low mood; and use A / B testing to continuously optimize the recommendation algorithm and anti-addiction strategy.

[0012] The working principle of the above technical solution is: comprehensively collect multi-source data related to users from multiple channels, including platform logs (recording user logins, game behaviors, etc.), user behavior records (such as clicks, purchases, etc.) and third-party data sources (such as social media data, market research reports, etc.). And perform pre-processing operations such as cleaning, deduplication, and formatting on the collected data; build a preliminary user portrait based on basic information such as user age, gender, and region; use clustering algorithms (such as K-means) to segment users and identify different user groups, such as heavy players, casual players, and social players; construct user behavior patterns by analyzing time series data of user behavior, including game duration, game type preferences, etc.; use deep learning models (such as LSTM, Transformer) to train user data to capture complex patterns and potential needs of user behavior; extract key features from user behavior data through feature engineering as input to the deep learning model; pre-process text data such as user comments and chat records, including segmentation words, part-of-speech tagging, etc., in order to conduct subsequent sentiment analysis; use sentiment dictionaries or sentiment classification models based on advanced technologies such as BERT to conduct sentiment analysis on user texts, and monitor users' emotional changes in real time, such as anger, disappointment, excitement, etc.; combine user portraits to analyze the reasons for user emotional changes, such as high game difficulty and poor social experience; based on user portraits and the prediction results of deep learning models, recommend games that meet users' preferences, while considering the user's emotional state to provide more considerate services; adjust the push strategy according to the user's emotional changes, such as avoiding pushing stimulating content when the user is depressed, in order to maintain the user's positive emotions; use A / B testing and other methods to continuously optimize the recommendation algorithm and anti-addiction strategy to improve recommendation effects and user satisfaction.

[0013] The effects of the above technical solutions are as follows: by obtaining data from multiple channels such as platform logs, user behavior records, and third-party data sources, it is possible to have a more comprehensive understanding of users and build a more accurate user portrait; by segmenting users based on clustering algorithms, it is helpful to identify the characteristics and needs of different user groups, and provide strong support for subsequent personalized recommendations and services; by combining user portraits and the prediction results of deep learning models, it is possible to recommend games that meet users' preferences, improve the accuracy of recommendations and user satisfaction; by analyzing the time series data of user behavior, it is possible to construct user behavior patterns, such as game duration, game type preferences, etc., to provide important references for subsequent recommendations and services; by using deep learning models to train user data, it is possible to capture the complex patterns and potential needs of user behavior and improve the intelligence level of recommendations; Using sentiment dictionaries or sentiment classification models to perform sentiment analysis on user texts can monitor users' emotional changes in real time, such as anger, disappointment, excitement, etc., providing a basis for subsequent adjustments to push strategies; analyzing the causes of emotional changes in combination with user portraits can help the platform better understand user needs and optimize game content and social experience; adjusting push strategies according to user emotional changes, such as avoiding pushing stimulating content when users are depressed, can help maintain users' positive emotions and improve user satisfaction; using A / B testing to continuously optimize recommendation algorithms and anti-addiction strategies can ensure the effectiveness and adaptability of strategies and continuously improve user experience; through accurate user portraits and personalized recommendations, it can ensure the accurate delivery of resources, improve operational efficiency and conversion rate; avoid unnecessary advertising and push, reduce resource waste, and reduce operating costs.

[0014] In one embodiment of the present invention, the S2 includes: S21. Use sensor technology and big data processing platforms to monitor server load, game popularity and user distribution in real time, and collect real-time data, such as CPU usage, memory occupancy, network bandwidth, etc.; perform real-time analysis on the collected real-time data to identify anomalies and trends; S22. Based on historical data, the ARIMA time series analysis model is used to predict the peak hours of the game, and the prediction results are revised based on external factors, including holidays, activities, and new version releases; S23. Based on the preset warning threshold, when the predicted value reaches or exceeds the threshold, the resource scheduling mechanism is triggered, and the server capacity is dynamically adjusted according to the prediction result, such as increasing or decreasing the number of servers; S24. Continuously optimize resource scheduling strategies through resource utilization efficiency evaluation indicators, such as CPU utilization, memory occupancy, etc.; regularly collect resource utilization data, analyze resource utilization efficiency through resource utilization data, adjust resource allocation ratios based on analysis results, and optimize server architecture.

[0015] The working principle of the above technical solution is: using sensor technology and big data processing platform to conduct all-weather, all-round real-time monitoring of server load, game popularity and user distribution. Including real-time tracking of key indicators such as CPU usage, memory occupancy, network bandwidth, etc.; in the process of real-time monitoring, the system will continuously collect real-time data of these key indicators to ensure the accuracy and completeness of the data. The system will immediately conduct in-depth analysis of the collected real-time data, including identifying anomalies in the data, such as a sudden surge in CPU usage or a sharp drop in memory occupancy, and analyzing trends in the data, such as gradual changes in user distribution or a continuous increase in game popularity; based on historical data, the system uses the ARIMA time series analysis model to accurately predict the peak hours of the game. The ARIMA model can capture the time series characteristics in the data, thereby providing reliable prediction results; considering that external factors such as holidays, activities and new version releases may have a significant impact on the peak hours of the game, the system will make necessary corrections to the prediction results of the ARIMA model. The system will preset a series of warning thresholds according to actual needs. When the predicted value reaches or exceeds these thresholds, the system will immediately trigger the resource scheduling mechanism; based on the prediction results, the system will dynamically adjust the server capacity. If it is predicted that the peak period of the game is about to arrive, the system will increase the number of servers in advance to ensure that the game can run smoothly. On the contrary, if it is predicted that the popularity of the game will decrease, the system will appropriately reduce the number of servers to save resources; the system continuously evaluates the resource scheduling strategy through a series of resource utilization efficiency evaluation indicators, such as CPU utilization, memory occupancy, etc. The system will regularly collect resource usage data and conduct in-depth analysis of this data. Through analysis, the system can find bottlenecks and problems in resource utilization, providing a basis for the formulation of optimization strategies; based on the analysis results, the system will adjust the resource allocation ratio. At the same time, the system will also optimize the server architecture, such as adding a cache layer, optimizing the network topology, etc., to further improve the overall performance and stability of the system.

[0016] The effects of the above technical solutions are as follows: through sensor technology and big data processing platform, server load, game popularity and user distribution can be monitored in real time, and key data such as CPU usage, memory occupancy, network bandwidth, etc. can be collected, which is helpful to timely discover system anomalies, such as overload, bottleneck or potential failure points; in-depth analysis of the collected real-time data can identify anomalies and trends in the data, warn of potential problems in advance, and provide sufficient time for the operation and maintenance team to intervene and deal with them, avoiding or reducing service interruptions; based on historical data, the ARIMA time series analysis model is used to predict the peak period of the game, understand the changing trend of game traffic in advance, and provide a scientific basis for resource scheduling; considering the impact of external factors such as holidays, activities and new version releases on game traffic, the prediction results are corrected, the accuracy of the prediction is improved, and the effectiveness of the resource scheduling strategy is ensured; according to the prediction results and the preset warning threshold, the resource scheduling mechanism is triggered to dynamically adjust the server capacity, which helps to balance resource supply and demand, ensure that the game can run stably during peak hours, and avoid resource waste; through resource utilization efficiency evaluation indicators such as CPU usage and memory occupancy, the resource scheduling strategy is continuously optimized. It helps to find and solve bottlenecks and problems in resource use and improve resource utilization efficiency; regularly collect resource use data and conduct in-depth analysis to help understand the actual situation of resource use and provide data support for the adjustment of resource allocation ratio; according to the analysis results, optimize the server architecture, such as adding a cache layer, optimizing the network topology, etc., to improve the overall performance and stability of the system and reduce operating costs; through real-time monitoring, accurate prediction and dynamic resource scheduling, ensure that the game can run stably during peak hours and provide a smooth gaming experience; through resource utilization efficiency optimization and server architecture optimization, reduce operating costs and improve resource utilization efficiency. It helps to invest more resources in game development and user experience improvement, and further improve user satisfaction.

[0017] In one embodiment of the present invention, the S23 includes: The prediction value of the peak gaming period output by the prediction module of the real-time monitoring system is compared with the preset warning threshold; the warning threshold is set comprehensively based on historical data, business needs, server performance and other factors, and is divided into different levels, such as slight warning, moderate warning and severe warning; if the prediction value reaches or exceeds the warning threshold, the resource scheduling preparation process is immediately triggered, and the warning event is recorded at the same time, including the trigger time, prediction value, warning level and other information; wherein, the prediction value is calculated by the following model:

[0018] in, represents the predicted value of the peak gaming period at time t; The weight coefficient representing the historical average activity; Indicates the historical average activity; represents the weight of the trend coefficient; represents the trend factor at time t; represents the smoothing coefficient of the trend factor; Indicates the weight of the special time impact coefficient; Indicates time The weight of the special event impact coefficient; Indicates time The impact value of a special event; n represents the time window size of the special event; represents the error term; Based on the prediction results and warning levels, specific resource scheduling strategies are formulated; for minor warnings, measures such as increasing virtual resources and optimizing existing resource allocation are taken; for moderate warnings, consideration is given to increasing the number of physical servers and adjusting the network architecture; for severe warnings, emergency resource allocation plans need to be initiated, including rapid deployment of backup server clusters and activation of off-site disaster recovery centers.

[0019] And consider external optimization resource scheduling strategies; According to the formulated resource scheduling strategy, resource scheduling operations are automatically performed, including increasing / decreasing the number of servers, adjusting network bandwidth, and configuring load balancing algorithms.

[0020] The working principle of the above technical solution is: monitor the output of the system prediction module in real time, compare the predicted value with the preset warning threshold, and trigger the corresponding resource scheduling strategy according to the comparison result. This process combines multiple links such as real-time data analysis, warning mechanism, resource scheduling strategy formulation and dynamic adjustment, and automatic execution of resource scheduling operations, aiming to ensure that the game platform can maintain stable operation when facing different load pressures; the system monitors the game peak period prediction value output by the prediction module in real time. Compare the predicted value with the preset warning threshold. The warning threshold is set comprehensively based on historical data, business needs, server performance and other factors, and is divided into different levels, such as slight warning, moderate warning and severe warning. If the predicted value reaches or exceeds the warning threshold, the system immediately triggers the resource scheduling preparation process. At the same time, the system records the warning event, including detailed information such as trigger time, predicted value, warning level, etc., for subsequent analysis and optimization; based on the prediction results and warning level, the system formulates a specific resource scheduling strategy.

[0021] For minor warnings, measures taken include increasing virtual resources, optimizing existing resource allocation, etc.

[0022] For moderate warnings, consider increasing the number of physical servers, adjusting the network architecture, etc.

[0023] For serious warnings, emergency resource allocation plans need to be initiated, such as quickly deploying backup server clusters and activating off-site disaster recovery centers.

[0024] The system also considers the impact of external factors (such as holidays, activities, and new version releases) on resource demand and dynamically adjusts the resource scheduling strategy. For example, during holidays, user activity is expected to increase, and the system needs to increase server capacity in advance; the initial release of a new version may be accompanied by a large influx of users, and the system needs to strengthen network bandwidth and load balancing capabilities. Optimize resource scheduling strategies based on intelligent algorithms (such as reinforcement learning and deep learning), continuously train models through historical data and real-time feedback, and improve the accuracy and efficiency of resource scheduling. According to the formulated resource scheduling strategy, the system automatically performs resource scheduling operations, including increasing / decreasing the number of servers, adjusting network bandwidth, and configuring load balancing algorithms; the system ensures that these operations can be performed quickly and accurately to cope with resource demands under different load pressures.

[0025] The effect of the above technical solution is: through real-time monitoring of the output of the system prediction module, it can accurately predict key indicators such as server load, game popularity and user distribution during the peak period of the game, and timely compare with the preset warning threshold, discover potential load pressure in advance, and provide a scientific basis for resource scheduling; once the predicted value reaches or exceeds the warning threshold, the system immediately triggers the resource scheduling preparation process and records the warning event. This rapid response mechanism can ensure that effective measures are taken in time before the load pressure increases to avoid service interruption or performance degradation; the warning threshold is set comprehensively based on historical data, business needs and server performance, and is divided into different levels (such as minor warning, moderate warning and severe warning), which helps to adopt different resource scheduling strategies according to the warning level to achieve flexible allocation and efficient use of resources; the system has formulated specific resource scheduling strategies for different degrees of load pressure. From increasing virtual resources and optimizing existing resource configurations to increasing the number of physical servers and adjusting the network architecture, and then to launching emergency resource allocation plans, the system can cope with various load pressures and ensure the stable operation of the game platform. The system not only formulates resource scheduling strategies based on prediction results and warning levels, but also considers the impact of external factors (such as holidays, activities, and new version releases) on resource demand, and dynamically adjusts the strategies, which helps to ensure that the resource scheduling strategies are always consistent with the actual needs of the game platform. The resource scheduling strategies are optimized based on intelligent algorithms (such as reinforcement learning and deep learning), and the models are continuously trained through historical data and real-time feedback to improve the accuracy and efficiency of resource scheduling and realize refined resource management. According to the formulated resource scheduling strategies, the system automatically performs resource scheduling operations, including increasing / decreasing the number of servers, adjusting network bandwidth, and configuring load balancing algorithms, which can reduce manual intervention and improve the efficiency and accuracy of resource scheduling. Through the organic combination of accurate prediction, graded warning, flexible scheduling, and intelligent algorithm optimization, efficient management of game platform resources is achieved, which helps to reduce operating costs, improve resource utilization, and provide strong support for the long-term and stable development of the game platform.The above formula combines the historical average activity, trend factors, special event impacts and error terms to provide a comprehensive prediction model to meet the prediction needs of game peak hours under different circumstances; the weight coefficients in the model can be adjusted according to actual conditions, so that the model can adapt to different business needs and market changes; by setting different warning thresholds, the model can predict possible peak hours in advance and trigger corresponding resource scheduling preparation processes, thereby improving the system's responsiveness and service quality; according to different warning levels, the model can guide the adoption of different resource scheduling strategies, such as increasing virtual resources, adjusting network architecture or enabling backup server clusters, thereby optimizing resource usage. The model can improve the efficiency and cost-effectiveness of use; through prediction and early warning, the model helps to reduce service interruptions or performance degradation caused by unforeseen peak periods, and improve the stability and reliability of the system; the model makes predictions based on historical data and real-time data, providing data support for decision makers, making decisions more scientific and accurate; the model structure allows the addition of more features and parameters to adapt to new situations that may arise in the future, such as new market trends or special events; the model can be integrated into the real-time monitoring system, update the predicted value in real time, and adjust resource allocation in time according to the predicted results to cope with the upcoming peak period; through prediction and early warning, it can avoid over-allocation of resources during non-peak hours, thereby reducing operating costs.

[0026] In one embodiment of the present invention, the S24 includes: S241. Compare the resource usage data collected in real time with the historical data, analyze the changing trend of resource usage, and identify potential bottlenecks and waste points; S242. Analyze the relationship between resource utilization efficiency and user experience based on user feedback and game experience data, and clarify the priority of resource optimization; S243. Predict resource usage data based on machine learning algorithms, identify fluctuations in resource demand in advance, conduct cost-benefit analysis based on resource utilization efficiency and business needs, and determine the optimal balance point for resource scheduling; S244. Based on the prediction results and cost-benefit analysis, formulate specific resource scheduling strategies, including server expansion / reduction plans, network bandwidth adjustment plans, load balancing strategy optimization, etc., and set up corresponding monitoring and evaluation mechanisms; S245. Based on the automated resource scheduling system, the resource scheduling operation is automatically triggered according to the preset policies and thresholds. After the resource scheduling is implemented, the resource usage is tracked through the real-time monitoring system, user feedback is collected, and potential problems are solved; S246. Dynamically adjust resource scheduling strategies based on monitoring results and user feedback, regularly evaluate server architecture, and identify and optimize potential bottlenecks and redundant parts.

[0027] The working principle of the above technical solution is: through the comparison of real-time collected resource usage data with historical data, correlation analysis of user feedback and game experience data, machine learning algorithm prediction, cost-benefit analysis and implementation of automated resource scheduling system, the optimization of resource utilization efficiency and dynamic adjustment of server architecture are realized; the system collects resource usage data in real time, such as CPU utilization, memory occupancy, network bandwidth, etc., compares these real-time data with historical data, analyzes the changing trend of resource utilization, and identifies potential bottlenecks and waste points through data analysis to provide a basis for subsequent resource optimization; combines user feedback and game experience data (such as delay, freeze, disconnection rate, etc.) to analyze the correlation between resource utilization efficiency and user experience; clarifies the priority of resource optimization to ensure that the resource problems that have the greatest impact on user experience are solved first; predicts resource usage data based on machine learning algorithms to identify fluctuations in resource demand in advance action; conduct cost-benefit analysis based on resource utilization efficiency and business needs to determine the optimal balance point for resource scheduling; formulate specific resource scheduling strategies based on prediction results and cost-benefit analysis, including server expansion / reduction plans, network bandwidth adjustment plans, load balancing strategy optimization, etc.; set up corresponding monitoring and evaluation mechanisms to track resource usage in real time and collect user feedback; based on the automated resource scheduling system, automatically trigger resource scheduling operations according to preset strategies and thresholds; after implementing resource scheduling, track resource usage through a real-time monitoring system; collect user feedback, promptly discover and resolve potential problems, and ensure the effectiveness of resource scheduling; dynamically adjust resource scheduling strategies based on monitoring results and user feedback; regularly evaluate server architecture, identify and optimize potential bottlenecks and redundant parts; improve the scalability and stability of the system to ensure that the game platform can maintain stable operation under various load pressures.

[0028] The effects of the above technical solutions are as follows: by comparing the resource usage data collected in real time with the historical data, it is possible to accurately analyze the changing trends of resource usage, timely identify potential bottlenecks and waste points, and provide a clear direction and basis for subsequent resource optimization; by combining user feedback and game experience data, an in-depth analysis of the relationship between resource usage efficiency and user experience can be conducted, which can clarify the priority of resource optimization, and help the game platform to prioritize solving the problems that have the greatest impact on user experience under limited resources, thereby improving the overall service quality and user satisfaction; by predicting resource usage data based on machine learning algorithms, it is possible to identify fluctuations in resource demand in advance, providing the game platform with sufficient time to prepare and adjust resources, and ensuring stable operation during peak hours or emergencies; on the basis of predicting resource demand, combined with resource utilization efficiency and business needs, Cost-benefit analysis can determine the optimal balance point for resource scheduling, which helps the game platform to minimize costs while meeting user needs and improve overall operational efficiency. Based on the automated resource scheduling system, resource scheduling operations are automatically triggered according to preset strategies and thresholds, reducing the frequency and difficulty of manual intervention and improving the efficiency and accuracy of resource scheduling. After implementing resource scheduling, tracking resource usage through a real-time monitoring system and collecting user feedback can help to promptly identify and resolve potential problems and ensure the effectiveness and stability of resource scheduling. Dynamically adjusting resource scheduling strategies based on monitoring results and user feedback can help the game platform maintain flexibility and competitiveness in a changing market environment. Regularly evaluating the server architecture, identifying and optimizing potential bottlenecks and redundant parts, and improving the scalability and stability of the system provide strong guarantees for the long-term development of the game platform.

[0029] In one embodiment of the present invention, the S243 includes: Clean the resource usage data collected in real time, extract features closely related to resource usage based on business scenarios, such as user activity, peak access time, business type distribution, etc., and construct feature vectors; select the most critical features for resource demand prediction based on correlation analysis; Use historical resource usage data to train the built-in machine learning model, adjust the model parameters through cross-validation, and evaluate the model's prediction accuracy and generalization ability by comparing the prediction results with actual resource usage; Based on the trained model, make short-term forecasts of resource demand in the near future (such as the next few hours or days), analyze the long-term trend of resource demand in combination with historical data, and predict the total resource demand in the future (such as monthly or quarterly); By comparing the difference between the predicted value and the actual value, the fluctuation of resource demand, including seasonal fluctuation, cyclical fluctuation and random fluctuation, etc., is identified. Based on the resource demand forecast results, the cost under different resource scheduling schemes is estimated, including server rental fees, network bandwidth fees, operation and maintenance costs, etc. The cost is estimated by the following formula:

[0030] in, represents the total cost, represents fixed cost, represents variable cost, represents the risk cost, represents opportunity cost; Fixed costs Obtained through the following formula:

[0031] in, represents the rental cost; represents the unit price of the i-th rental resource (e.g., office rent, equipment rental fee); Indicates the number of i-th leased resources; represents the licensing cost; represents the unit price of the jth software or service license; represents the number of the jth software or service license; represents salary cost; represents the average hourly wage of the kth employee or position; represents the working hours of the kth employee or position; represents indirect costs; It represents fixed indirect costs, such as management expenses, which do not change with production volume;

[0032] Variable costs Obtained through the following formula:

[0033] in, represents the computational cost; represents the unit price of the i-th computing resource (such as server, cloud instance); Indicates the usage of the i-th computing resource (such as CPU hours); represents the usage time of the i-th computing resource; represents bandwidth cost; Indicates the unit price of the jth bandwidth resource (such as network traffic); Indicates the usage of the jth bandwidth resource (such as GB or TB); represents the usage time of the jth bandwidth resource; represents the storage cost; represents the unit price of the kth storage resource (such as hard disk, cloud storage); Indicates the usage of the kth storage resource (such as GB or TB); Indicates the usage time of the kth storage resource; Indicates operation and maintenance cost; Indicates the unit price of the lth operation and maintenance service (such as system monitoring and backup); Indicates the usage of the lth operation and maintenance service (such as monitoring hours); Indicates the usage time of the lth operation and maintenance service;

[0034] Risk Cost Obtained through the following formula:

[0035] in, represents the cost of downtime; Indicates the probability of downtime occurring; represents the expected probability of downtime event; Indicates the average loss when a downtime event occurs (e.g., lost revenue, repair cost); represents the security cost; Indicates the probability of a security incident occurring; Represents the expected probability of a security incident; Represents the average loss when a security incident occurs (such as data leakage, system repair costs);

[0036] The opportunity cost Obtained through the following formula:

[0037] in, represents a potential loss of income; Indicates the probability of failing to seize market opportunities; Represents the expected probability of failing to seize a market opportunity; Represents the average potential revenue loss when market opportunities are not seized; Indicates loss of market share; represents the probability of market share loss; represents the expected probability of market share loss; Represents the average loss in the event of a market share loss (i.e., long-term revenue loss due to a loss of market share);

[0038] In the above formula, represents the unit price of the yth resource or service; , , , , , , They represent the number of servers, number of software licenses, employee hours, computing resource usage, bandwidth usage, storage usage, and operation and maintenance workload respectively; , , , Indicates the usage time of the corresponding resource; It represents the expected probability of the risk event occurring; Indicates the average loss caused by risk events; N, M, O, and P respectively represent the number of different resources or service categories; Based on business needs, user experience and market competition, evaluate the benefits of different resource scheduling solutions in terms of business growth, improved user satisfaction, etc.

[0039] Based on cost-benefit analysis, determine the optimal balance point for resource scheduling and formulate specific resource scheduling strategies, such as server expansion / reduction plans, network bandwidth adjustment plans, etc.

[0040] The working principle of the above technical solution is as follows: real-time collection of resource usage data, such as CPU usage, memory occupancy, network bandwidth, etc.; cleaning of the collected data, removal of outliers and duplicate data, and extraction of features closely related to resource usage based on business scenarios, such as user activity, peak access time, business type distribution, etc., and construction of feature vectors; using correlation analysis to screen out the most critical features for resource demand prediction, reduce model complexity, and improve prediction efficiency; using historical resource usage data to train built-in machine learning models (such as time series prediction models, regression models, etc.); adjusting model parameters through cross-validation to optimize model performance; comparing prediction results with actual resource usage to evaluate the prediction accuracy and generalization ability of the model; based on the trained model, making short-term predictions on resource demand in the near future (such as the next few hours or days); combining historical data to analyze the long-term trend of resource demand and predict Predict the total resource demand in the future (such as monthly or quarterly); identify the fluctuation of resource demand by comparing the difference between the predicted value and the actual value, including seasonal fluctuation, cyclical fluctuation and random fluctuation; based on the resource demand forecast results, estimate the cost under different resource scheduling schemes, including server rental fees, network bandwidth fees, operation and maintenance costs, etc.; evaluate the benefits of business growth and user satisfaction improvement brought by different resource scheduling schemes in combination with business needs, user experience and market competition; comprehensively consider costs and benefits to determine the best balance point for resource scheduling; formulate specific resource scheduling strategies based on the results of cost-benefit analysis, such as server expansion / reduction plans, network bandwidth adjustment plans, etc.; consider the elasticity and scalability of resource use to ensure that the strategy can adapt to the needs of future business development; set up corresponding monitoring and evaluation mechanisms to track resource usage in real time and adjust resource scheduling strategies in a timely manner.

[0041] The effects of the above technical solutions are as follows: by cleaning the resource usage data collected in real time, extracting key features based on business scenarios, and constructing feature vectors, a high-quality data foundation is provided for subsequent machine learning model training; the most critical features for resource demand prediction are screened out, reducing model complexity and improving prediction accuracy; the machine learning model is trained using historical resource usage data, and the model parameters are adjusted through cross-validation to ensure the prediction accuracy and generalization ability of the model. Based on the trained model, short-term prediction of recent resource demand can be achieved, and the long-term change trend can be analyzed in combination with historical data to predict the total resource demand in the future, providing a scientific basis for resource scheduling; by comparing the difference between the predicted value and the actual value, seasonal fluctuations, cyclical fluctuations and random fluctuations in resource demand can be identified, providing a more detailed reference for resource scheduling; the costs under different resource scheduling schemes are estimated, including server rental fees, network bandwidth fees, operation and maintenance costs, etc., and the benefits are evaluated in combination with business needs, user experience and market competition, ensuring the economy and effectiveness of resource scheduling; based on cost-benefit analysis, resource scheduling can be determined The best balance point is determined by formulating specific resource scheduling strategies, such as server expansion / reduction plans, network bandwidth adjustment plans, etc., which improves the efficiency and flexibility of resource use; when formulating resource scheduling strategies, the elasticity and scalability of resource use are considered to ensure that the strategy can adapt to the needs of future business development and improve the competitiveness and market response speed of the game platform; by accurately predicting resource demand and optimizing resource scheduling strategies, the game platform can maintain stable operation during peak hours or emergencies, reduce delays, freezes and other problems, and improve user experience and satisfaction; by optimizing resource scheduling strategies, unnecessary resource waste and operating costs can be reduced, and the profitability of the game platform can be improved. By dividing the total cost into fixed cost, variable cost, risk cost and opportunity cost, we can clearly understand the various expenses under different resource scheduling plans and help enterprises identify the composition of various expenses. This segmentation helps to more accurately predict and control resource inputs and avoid blind expansion or waste; by accurately estimating the costs of different resources (such as servers, bandwidth, storage, operation and maintenance, etc.), it can help enterprises evaluate the advantages and disadvantages of different scheduling plans and formulate more reasonable plans such as expansion / reduction and bandwidth adjustment to maximize resource utilization and reduce costs; taking risk costs and opportunity costs into consideration can help enterprises identify potential business risks (such as system downtime, security vulnerabilities, etc.) and quantitatively evaluate the impact of these risks on costs. In addition, the calculation of opportunity costs can help enterprises weigh market opportunities and avoid missing potential growth opportunities; by combining efficiency indicators such as business growth and improved user satisfaction, cost-benefit analysis can not only help enterprises control costs, but also find the best balance point in resource scheduling to maximize business goals.Enterprises can achieve better market competitiveness and user experience without exceeding their budgets; seasonal fluctuations, cyclical fluctuations, random fluctuations and other factors involved in the above formula can help enterprises predict changes in resource demand and make dynamic adjustments. This capability makes resource scheduling more flexible and efficient, and can respond to market changes and sudden demands; through precise resource scheduling and cost control, enterprises can improve overall operational efficiency, reduce ineffective resource input, and reduce waste, thereby providing guarantees for the sustainable development of enterprises; through quantitative cost models, enterprise management can clearly understand the input and return of various resources, enhance the transparency and scientific nature of decision-making, avoid making decisions based on intuition or experience, and thus improve the accuracy of decisions.

[0042] In one embodiment of the present invention, S3 includes: S31. Through natural language processing and image recognition technology, conduct real-time review of game content, identify and filter out harmful information, and establish a sensitive word library and image feature library for continuous updating and improvement; S32. Detect abnormal behaviors such as cheating and malicious attacks through AI algorithms, and discover and handle abnormal behaviors through behavioral pattern analysis and network traffic monitoring; S33. Based on the abnormal behavior database, the detection model is trained and optimized, and through the early warning mechanism, when abnormal behavior is detected, the defense measures are automatically triggered; S34. Take defensive measures based on the detection results, such as banning accounts, restricting IP access, strengthening security verification, etc., and regularly update and upgrade the security protection system; S35. Combine user feedback and automatic learning mechanisms to continuously optimize the audit model, use generative adversarial networks to generate simulated attack data, and enhance the robustness of the security protection system.

[0043] The working principle of the above technical solution is as follows: using natural language processing and image recognition technology to conduct real-time analysis of game content, identify and filter negative information, such as violence, pornography, gambling, etc.; establishing and continuously updating sensitive word libraries and image feature libraries, which contain known negative information features to assist in identifying new negative content; with the emergence of new negative information, the sensitive word library and image feature library will be continuously updated and improved to ensure the accuracy and comprehensiveness of the review; analyzing the behavior patterns of game users through AI algorithms to identify abnormal behaviors such as cheating and malicious attacks; analyzing user behavior patterns, such as login frequency, operating habits, etc., and comparing them with normal behavior patterns to find abnormalities; monitoring the network traffic of the game platform, identifying abnormal network activities, such as large amounts of data transmission, abnormal requests, etc., as an auxiliary basis for judging abnormal behavior; establishing an abnormal behavior database to store known abnormal behaviors The system uses samples to train and optimize detection models; uses an abnormal behavior database to train detection models to improve the model's ability to identify abnormal behaviors; when abnormal behaviors are detected, the early warning mechanism is automatically triggered to notify the security team to handle them; according to the detection results, corresponding defense measures are taken, such as banning accounts, restricting IP access, and strengthening security verification; the security protection system is regularly updated and upgraded to cope with new security threats; an emergency response mechanism is established to enable rapid response and handling when security incidents occur to reduce losses; the audit model is continuously optimized in combination with user feedback and automatic learning mechanisms to improve the accuracy and adaptability of the model; a generative adversarial network (GAN) is used to generate simulated attack data, which is used to train the security protection system and enhance its robustness; through continuous learning and optimization, the security protection system will be more intelligent and efficient, and will be able to better cope with various security threats.

[0044] The effects of the above technical solutions are as follows: through natural language processing and image recognition technology, real-time review of game content can be achieved, and bad information such as violence, pornography, gambling, etc. can be discovered and filtered in time, effectively improving the review efficiency; timely filtering of bad information can maintain a good atmosphere of the game platform, ensure the player's gaming experience, and reduce user loss caused by bad information; by continuously updating and improving sensitive word libraries and image feature libraries, the recognition rate of bad information can be continuously improved to adapt to the ever-changing information environment; through AI algorithms, abnormal behaviors such as cheating and malicious attacks can be detected and handled in time to improve the security of the game platform; through behavioral pattern analysis and network traffic monitoring methods, abnormal behaviors can be identified more accurately and false alarms and missed alarms can be reduced; when abnormal behavior is detected, defense measures can be automatically triggered, such as banning accounts, restricting IP access, etc., to prevent the spread of security threats in time; based on the abnormal behavior database, the detection model can be trained and optimization to improve the model's recognition accuracy of abnormal behavior; through the early warning mechanism, the security team can be notified in time when abnormal behavior is detected, providing a time window for rapid response and processing; a variety of defense measures are taken according to the detection results, such as banning accounts, restricting IP access, strengthening security verification, etc., to effectively deal with different types of security threats; regularly updating and upgrading the security protection system can improve the security and stability of the system and adapt to new security threats; through the emergency response mechanism, it can quickly respond to security incidents, reduce losses, and ensure the normal operation of the game platform; by combining user feedback and automatic learning mechanisms, the audit model is continuously optimized to improve the accuracy and adaptability of the model; using generative adversarial networks to generate simulated attack data, the robustness of the security protection system is enhanced, so that it can better deal with complex and changing security threats; through continuous learning and optimization, the performance and stability of the security protection system can be continuously improved, providing long-term protection for the safe operation of the game platform.

[0045] In one embodiment of the present invention, the S31 includes: Preprocess the text information in the game, including removing stop words, punctuation marks, special characters, etc. Based on the established sensitive word library, quickly match the preprocessed text to identify content containing sensitive words; For texts that do not directly match sensitive words, natural language processing technology is used to perform semantic analysis to identify the text's potential intent and emotional tendency; Preprocess the images in the game, including scaling, denoising, contrast enhancement, etc. Based on the deep learning algorithm, extract features from the preprocessed images and build an image feature library; The extracted features are compared with the features in the image feature library to identify bad images and classify them. At the same time, the generative adversarial network (GAN) is used to generate simulated bad images for training and optimizing the image recognition model. Based on the results of natural language processing and image recognition, intelligent filtering rules are formulated to automatically filter out the identified bad information; For cases where there may be misjudgments, a manual review mechanism is established; Establish user reporting channels to collect user reports and feedback on bad information and further optimize filtering strategies and models; Using machine learning algorithms, new sensitive words and image features discovered during the filtering process are automatically learned and updated to the sensitive word library and image feature library.

[0046] The working principle of the above technical solution is as follows: First, pre-process the text information in the game, including removing irrelevant information such as stop words, punctuation marks, special characters, etc., to reduce the complexity of subsequent processing; then, based on the established sensitive word library, quickly match the pre-processed text to identify the content containing sensitive words. The sensitive word library needs to be updated regularly to respond to new bad information and ensure the accuracy and timeliness of the review; for texts that are not directly matched to sensitive words, semantic analysis is performed through natural language processing technology to identify the potential intentions and emotional tendencies of the text; combined with the results of semantic analysis, potential bad information, such as implicit violence, pornography or gambling content, is discovered to ensure the comprehensiveness of the review; pre-process the images in the game, including operations such as scaling, denoising, and contrast enhancement to improve the accuracy of image recognition; based on the deep learning algorithm, feature extraction is performed on the pre-processed images to build an image feature library. The feature library needs to be continuously updated to cover more types of inappropriate images and enhance the recognition ability of the model; compare the extracted image features with the features in the image feature library to identify inappropriate images; classify the identified inappropriate images, such as by violence, pornography, gambling, etc., so that targeted filtering measures can be taken later; use generative adversarial networks (GANs) to generate simulated inappropriate images, which can be used to train and optimize image recognition models to improve the recognition accuracy and generalization ability of the models; through continuous iterative training, the image recognition model can better identify and process various types of inappropriate images; formulate intelligent filtering rules based on the results of natural language processing and image recognition, and use the formulated filtering rules to filter text and images in the game Automatically filter content to ensure that negative information is not spread; for situations where there may be misjudgments, set up a manual review mechanism, and have professional reviewers review the results of automatic filtering to ensure the accuracy of filtering; based on the results of manual review, continuously optimize filtering strategies and models to improve the accuracy and efficiency of review; establish user reporting channels to facilitate users to report and feedback negative information; collect user reports and feedback on negative information to further optimize filtering strategies and models, and improve the pertinence and effectiveness of review; use machine learning algorithms to automatically learn new sensitive words and image features discovered during the filtering process; update the learned new sensitive words and image features to the sensitive word library and image feature library to continuously improve the recognition ability of the review model.

[0047] The effect of the above technical solution is: through the rapid matching of pre-processed text information and sensitive word library, the content containing sensitive words can be quickly identified. At the same time, for texts that are not directly matched to sensitive words, semantic analysis using natural language processing technology can deeply explore the potential intentions and emotional tendencies of the text, thereby discovering potential negative information and ensuring the efficiency and accuracy of content review. After preprocessing the images in the game, deep learning algorithms are used to extract features and compare them with the features in the image feature library, which can accurately identify negative images, not only improving the review efficiency, but also enhancing the accuracy of recognition. Based on the results of natural language processing and image recognition, intelligent filtering rules are formulated to automatically filter the identified negative information, greatly reducing the burden of manual review and improving the efficiency of review. For situations where there may be misjudgments, a manual review mechanism is established to ensure the accuracy of filtering. The review method that combines manual and intelligent methods not only ensures the efficiency of the review, but also ensures the accuracy of the review. The sensitive word library and image feature library need to be updated regularly to cope with new negative information. The continuous update mechanism ensures that the review model can adapt to the ever-changing information environment and improve the timeliness and accuracy of recognition. The machine learning algorithm is used to automatically learn the new sensitive words and image features discovered during the filtering process and update them to the sensitive word library and image feature library. This automatic learning mechanism can continuously improve the recognition ability of the audit model, making it more intelligent and efficient; establish a user reporting channel to collect user reports and feedback on bad information. This user participation method can timely discover and deal with bad information in the game, improve the user experience and security of the game platform; according to user reports and feedback, further optimize the filtering strategy and model, and the optimization mechanism based on user feedback can continuously improve the accuracy and adaptability of the audit model, making it more in line with user needs and expectations; through efficient and accurate content review and intelligent filtering mechanism.

[0048] In one embodiment of the present invention, the S4 includes: S41. Based on natural language processing and GPT models, provide 24 / 7 intelligent customer service to handle user issues, such as game strategies, recharge issues, account recovery, etc.

[0049] S42. Introduce multi-round dialogue technology to improve the interactivity and satisfaction of customer service, regularly analyze customer service data, and optimize response techniques and strategies; S43. Adjust the response strategy according to the user's emotions, such as providing soothing responses to anger, and introduce an emotion recognition model to monitor the user's emotional changes in real time; S44. Provide personalized emotional conversation services based on user portraits and historical conversation records, regularly analyze emotional conversation data, and optimize emotion recognition models and response strategies; S45. Identify and guide positive topics through data analysis algorithms, push community activities and hot topic discussions based on user interests, and analyze user behaviors, such as likes, comments, and shares, and evaluate community activity based on the analysis results; S46. Regularly analyze community data, optimize community management strategies based on the analysis results, and continuously optimize customer service strategies and community management strategies based on user feedback.

[0050] The working principle of the above technical solution is as follows: users submit questions through the customer service portal in the game or on the official platform, such as game strategy inquiries, recharge consultations, account recovery requests, etc.; use natural language processing technology to parse the input text and extract key information; input the parsed information into the pre-trained GPT model to generate natural and fluent replies. The GPT model can provide targeted answers or suggestions based on contextual information; present the generated replies to users, and record customer service conversation data at the same time to provide a basis for subsequent analysis and optimization; introduce multi-round conversation technology to allow users to interact continuously with intelligent customer service until the problem is solved, thereby enhancing the interactivity and satisfaction of customer service; regularly mine and analyze customer service data to identify key indicators such as common user problems, response efficiency and satisfaction; based on the analysis results, adjust the reply words and strategies to improve the accuracy of the reply and user satisfaction; introduce an emotion recognition model to monitor the emotional changes in user conversations in real time, such as anger, anxiety, etc.; dynamically adjust the reply strategy according to user emotions. For example, for angry users, soothing replies are provided to ease their emotions; the effects of emotion recognition and response strategy adjustments are recorded for subsequent model optimization and training; user portraits are constructed based on user historical behaviors, preferences, and conversation records; based on user portraits, personalized emotional conversation services are provided to enhance user experience and stickiness; emotional conversation data is regularly analyzed to evaluate the effectiveness of personalized services, and emotion recognition models and response strategies are iteratively optimized; data analysis algorithms are applied to identify user interests and hot topics; community activities and hot topic discussions are pushed based on user interests to guide users to actively participate; user interaction behaviors such as likes, comments, and shares are monitored and analyzed to evaluate community activity; based on the analysis results, community activity is quantitatively evaluated to provide a basis for the adjustment of community management strategies; community data is regularly analyzed in depth to identify potential problems and challenges; based on the analysis results, community management strategies are adjusted, such as strengthening content review and optimizing user interaction rules; user feedback is collected and integrated as an important basis for optimizing customer service strategies and community management strategies; a closed-loop optimization mechanism is formed to continuously iterate and optimize customer service and community management strategies based on user feedback and data analysis results to improve user experience and satisfaction.

[0051] The effects of the above technical solutions are as follows: intelligent customer service based on natural language processing and GPT model can provide users with instant responses around the clock, solve various problems such as game strategies, recharge problems, account retrieval, etc., greatly improving service efficiency; the introduction of intelligent customer service effectively reduces the workload of manual customer service, allowing manual customer service to focus on handling more complex or special problems, improving the overall service quality; by introducing multi-round dialogue technology, intelligent customer service can interact with users continuously and in-depth, better understand user needs, and provide more accurate answers, thereby improving the interactivity of customer service and user satisfaction; regular analysis of customer service data and optimization of reply words and strategies can continuously improve the intelligence level and user experience of intelligent customer service; adjusting the reply strategy according to user emotions, such as providing soothing replies to anger, can effectively alleviate user negative emotions and improve user satisfaction. The introduction of the emotion recognition model has realized real-time monitoring of user emotions, providing strong support for emotional services; combining user portraits and historical conversation records to provide personalized emotional conversation services can enhance users' sense of belonging and loyalty and further improve user experience; identifying and guiding positive topics through data analysis algorithms, pushing community activities and hot topic discussions based on user interests can effectively stimulate user participation enthusiasm and increase community activity; analyzing user interactive behaviors such as likes, comments, and shares can deeply understand user preferences and behavioral habits, providing a strong basis for optimizing community management strategies; regularly analyzing community data and optimizing community management strategies based on the analysis results can continuously improve the pertinence and effectiveness of community management; continuously optimizing customer service strategies and community management strategies based on user feedback can ensure that services and management are always close to user needs and improve user satisfaction and loyalty.

[0052] One embodiment of the present invention, as Figure 2 As shown, a game platform management system based on artificial intelligence, the system includes: Data acquisition module: obtain multi-source data through multiple channels, build user portraits based on the obtained multi-source data, and train user data through deep learning models to identify user preferences, habits and potential needs; combine natural language processing technology to conduct sentiment analysis on user comments and chat records, identify user emotional changes, and provide accurate basis for personalized recommendations and anti-addiction strategies based on the analysis results, such as pushing games according to user preferences and adjusting push strategies according to emotional changes; Real-time monitoring module: monitors server load, game popularity and user distribution in real time. Based on historical data, it uses time series analysis to predict the peak hours of games. According to the prediction results, it dynamically adjusts resource allocation, such as increasing server capacity and optimizing network bandwidth to ensure smooth game operation. Anomaly detection module: Through natural language processing and image recognition technology, it conducts real-time review of game content, identifies and filters bad information, and uses AI algorithms to detect abnormal behaviors such as cheating and malicious attacks, and takes defensive measures based on the detection results; Strategy Adjustment Module: Based on natural language processing and GPT models, it provides 24 / 7 intelligent customer service, adjusts response strategies according to user emotions, such as providing soothing responses to anger, and identifies and guides positive topics based on data analysis algorithms, such as pushing community activities based on user interests to promote community interaction, and continuously optimizes customer service strategies and community management strategies based on user feedback.

[0053] The working principle of the above technical solution is as follows: First, the platform obtains multi-source data through multiple channels (such as user registration information, game behavior records, social media links, etc.); through the obtained multi-source data, a detailed user portrait is constructed, including the user's age, gender, game preferences, consumption habits, etc.; through the deep learning model, the user data is trained to identify the user's preferences, habits and potential needs. For example, by analyzing the user's historical game records, the user's favorite game type can be inferred; combined with natural language processing technology, sentiment analysis is performed on user comments and chat records to identify user emotional changes, such as positive, negative or neutral; based on the above analysis results, personalized game recommendations are provided to users, such as pushing new games according to user preferences. At the same time, the push strategy is adjusted according to the user's emotional changes to avoid pushing content that may cause discomfort when the user is in a low mood. In addition, the platform can also use these analyses to formulate anti-addiction strategies, such as sending reminders when the user's game time is too long; the platform monitors the server load, game popularity and user distribution in real time to obtain the latest game operation status; based on historical data, time series analysis technology is used to predict the peak hours of the game. For example, by analyzing the user login data of the past few months, it is possible to predict which time periods are the most frequent for users to log in; based on the prediction results, the platform dynamically adjusts resource allocation. Before the peak period arrives, the server capacity is increased, the network bandwidth is optimized, etc., to ensure that the game can run smoothly and avoid problems such as freezing or disconnection caused by insufficient resources; through natural language processing and image recognition technology, the platform conducts real-time review of game content, including identifying and filtering bad information, such as violence, pornography, etc.; using AI algorithms to detect abnormal user behavior, such as cheating, malicious attacks, etc. The user's behavior pattern is analyzed through AI algorithms to identify abnormal behaviors that are inconsistent with normal behavior; based on the detection results, the platform takes corresponding defensive measures. For example, for cheating behavior, the cheating account can be banned; for malicious attacks, security measures can be taken to protect the safety of other users; based on natural language processing and GPT models, the platform provides 24 / 7 intelligent customer service, and the intelligent customer service can adjust the response strategy according to the user's emotions, such as providing soothing responses to anger. Using data analysis algorithms, intelligent customer service can also identify and guide positive topics. For example, relevant community activities are pushed based on user interests to promote interaction and communication between users; based on user feedback and interaction data, the platform continuously optimizes customer service strategies and community management strategies, including improving the response quality of intelligent customer service and increasing the attractiveness of community activities.

[0054] The effects of the above technical solutions are as follows: through the training of user data by deep learning models, it is possible to accurately identify user preferences, habits and potential needs, thereby providing users with personalized game recommendations, increasing the possibility of users finding games of interest, and improving user satisfaction and loyalty; combining natural language processing technology for sentiment analysis, it is possible to capture user emotional changes in real time, and adjust the push strategy according to the emotion, avoiding pushing content that may cause discomfort when the user is in a low mood, while increasing the chance of pushing relevant content when the user is in a positive mood, further improving the user experience; through real-time monitoring of server load, game popularity and user distribution, and using time series analysis to predict the peak hours of the game, the platform can make resource adjustment decisions in advance, ensuring that the game can run smoothly during peak hours, avoiding problems such as freezes or disconnections due to insufficient resources, and improving operational efficiency; dynamic adjustment of resource allocation also means that resource investment can be reasonably reduced during low periods, thereby optimizing resources. Utilize and reduce costs; through natural language processing and image recognition technology, the platform can conduct real-time review of game content, effectively identify and filter out bad information, help maintain a healthy and positive gaming environment, and protect users from bad information; use AI algorithms to detect abnormal behaviors such as cheating and malicious attacks, and take defensive measures based on the detection results, which can effectively curb the occurrence of bad behaviors and ensure the fairness and security of the game; intelligent customer service based on natural language processing and GPT models can provide 24-hour customer service support and adjust response strategies according to user emotions, which improves customer service efficiency and enhances user satisfaction and trust; identify and guide positive topics through data analysis algorithms, such as pushing community activities based on user interests, which can promote interaction and communication between users and enhance the cohesion and activity of the community; continuously optimize customer service strategies and community management strategies based on user feedback to ensure that the platform always stays at the forefront of user needs and continuously improves user experience.

[0055] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A game platform management method based on artificial intelligence, characterized in that: The method comprises: S1. Obtain multi-source data through multiple channels, build user portraits based on the acquired multi-source data, and train user data through deep learning models to identify user preferences, habits and potential needs; S2. Monitor server load, game popularity and user distribution in real time. Based on historical data, use time series analysis to predict game peak hours and dynamically adjust resource allocation according to the prediction results. S3. Use natural language processing and image recognition technology to conduct real-time review of game content, identify and filter bad information, and use AI algorithms to detect abnormal behavior and take defensive measures based on the detection results; S4. Based on natural language processing and GPT models, provide 24 / 7 intelligent customer service and adjust the response strategy according to user emotions.

2. According to the game platform management method based on artificial intelligence in claim 1, it is characterized in that: Said S1 comprises: S11, acquiring multi-source data from multiple channels, wherein the multiple channels include platform logs, user behavior records, and third-party data sources, and preprocessing the acquired multi-source data; S12. Build a preliminary user portrait based on basic user information, game preferences and consumption habits; and segment users based on clustering algorithms to identify different user groups; S13. Combine user behavior time series analysis to build user behavior patterns and train user data through deep learning models; S14. Through feature engineering, extract key features of user behavior as model input and perform cross-validation. Pre-process user comments and chat records through natural language processing technology. S15. Perform sentiment analysis on user texts through sentiment dictionaries or sentiment classification models to monitor user sentiment changes in real time, such as anger, disappointment, and excitement; S16. Analyze the reasons for emotional changes based on user portraits, and recommend games to users based on user portraits and deep learning model prediction results; S17. Adjust the push strategy according to changes in user emotions, and use A / B testing to continuously optimize the recommendation algorithm and anti-addiction strategy.

3. The game platform management method based on artificial intelligence according to claim 1, characterized in that: The S2 comprises: S21. Use sensor technology and big data processing platforms to monitor server load, game popularity and user distribution in real time, collect real-time data, analyze the collected real-time data in real time, and identify abnormal points and trends; S22. Based on historical data, the ARIMA time series analysis model is used to predict the peak gaming hours, and the prediction results are revised based on external factors; S23, based on the preset warning threshold, when the predicted value reaches or exceeds the threshold, trigger the resource scheduling mechanism, and dynamically adjust the server capacity according to the prediction result; S24. Continuously optimize resource scheduling strategies through resource utilization efficiency evaluation indicators; regularly collect resource utilization data, analyze resource utilization efficiency through resource utilization data, adjust resource allocation ratios based on analysis results, and optimize server architecture.

4. The game platform management method based on artificial intelligence according to claim 3 is characterized in that: The S23 comprises: Real-time monitoring of the game peak period forecast value output by the forecast module of the system, and comparing the forecast value with the preset warning threshold; if the forecast value reaches or exceeds the warning threshold, the resource scheduling preparation process is immediately triggered, and the warning event is recorded at the same time; Develop resource scheduling strategies based on forecast results and warning levels; It also considers the impact of external factors on resource demand and dynamically adjusts resource scheduling strategies; and optimizes resource scheduling strategies based on machine learning algorithms; According to the formulated resource scheduling strategy, resource scheduling operations are automatically performed, including increasing / decreasing the number of servers, adjusting network bandwidth, and configuring load balancing algorithms.

5. The game platform management method based on artificial intelligence according to claim 3 is characterized in that: The S24 comprises: S241. Compare the resource usage data collected in real time with the historical data, analyze the changing trend of resource usage, and identify potential bottlenecks and waste points; S242. Analyze the relationship between resource utilization efficiency and user experience based on user feedback and game experience data, and clarify the priority of resource optimization; S243. Predict resource usage data based on machine learning algorithms, identify fluctuations in resource demand, conduct cost-benefit analysis based on resource usage efficiency and business needs, and determine the balance point of resource scheduling; S244. Develop resource scheduling strategies based on forecast results and cost-benefit analysis, and set up corresponding monitoring and evaluation mechanisms; S245. Based on the automated resource scheduling system, the resource scheduling operation is automatically triggered according to the preset policies and thresholds. After the resource scheduling is implemented, the resource usage is tracked through the real-time monitoring system, user feedback is collected, and potential problems are solved; S246. Dynamically adjust resource scheduling strategies based on monitoring results and user feedback, regularly evaluate server architecture, and identify and optimize potential bottlenecks and redundant parts.

6. The game platform management method based on artificial intelligence according to claim 5, characterized in that: The S243 includes: Clean the resource usage data collected in real time, extract features closely related to resource usage based on business scenarios, and construct feature vectors; Use historical resource usage data to train the built-in machine learning model, adjust the model parameters through cross-validation, and evaluate the model's prediction accuracy and generalization ability by comparing the prediction results with actual resource usage; Based on the trained model, make short-term forecasts of recent resource demand, combine historical data to analyze the long-term trend of resource demand, and predict the total resource demand in the future. By comparing the difference between the predicted value and the actual value, the fluctuation of resource demand can be identified, and the cost under different resource scheduling schemes can be estimated based on the resource demand forecast results; Based on cost-benefit analysis, determine the optimal balance point for resource scheduling and formulate specific resource scheduling strategies.

7. The game platform management method based on artificial intelligence according to claim 1, characterized in that: The S3 includes: S31. Through natural language processing and image recognition technology, conduct real-time review of game content, identify and filter out harmful information, and establish a sensitive word library and image feature library for continuous updating and improvement; S32. Detect abnormal behaviors through AI algorithms, and discover and handle abnormal behaviors through behavioral pattern analysis and network traffic monitoring; S33. Based on the abnormal behavior database, the detection model is trained and optimized, and through the early warning mechanism, when abnormal behavior is detected, the defense measures are automatically triggered; S34. Take defensive measures based on the test results and regularly update and upgrade the security protection system; S35. Combine user feedback and automatic learning mechanisms to continuously optimize the audit model, use generative adversarial networks to generate simulated attack data, and enhance the robustness of the security protection system.

8. The game platform management method based on artificial intelligence according to claim 7, characterized in that: The S31 includes: Preprocess the text information in the game, and quickly match the preprocessed text based on the established sensitive word library to identify content containing sensitive words; For texts that do not directly match sensitive words, natural language processing technology is used to perform semantic analysis to identify the text's potential intent and emotional tendency; Preprocess the images in the game, including scaling, denoising, contrast enhancement, etc. Based on the deep learning algorithm, extract features from the preprocessed images and build an image feature library; The extracted features are compared with the features in the image feature library to identify bad images and classify them. At the same time, the generative adversarial network (GAN) is used to generate simulated bad images for training and optimizing the image recognition model. Based on the results of natural language processing and image recognition, intelligent filtering rules are formulated to automatically filter out the identified bad information; Using machine learning algorithms, new sensitive words and image features discovered during the filtering process are automatically learned and updated to the sensitive word library and image feature library.

9. The game platform management method based on artificial intelligence according to claim 1, characterized in that: The S4 comprises: S41. Based on natural language processing and GPT models, provide 24 / 7 intelligent customer service to handle user issues; S42. Introduce multi-round dialogue technology to improve the interactivity and satisfaction of customer service, regularly analyze customer service data, and optimize response techniques and strategies; S43, adjusting the response strategy according to the user's emotions, and introducing an emotion recognition model to monitor the user's emotional changes in real time; S44. Provide personalized emotional conversation services based on user portraits and historical conversation records, regularly analyze emotional conversation data, and optimize emotion recognition models and response strategies; S45. Identify and guide positive topics through data analysis algorithms, push community activities and hot topic discussions based on user interests, analyze user behavior, and evaluate community activity based on the analysis results; S46. Regularly analyze community data, optimize community management strategies based on the analysis results, and continuously optimize customer service strategies and community management strategies based on user feedback.

10. A game platform management system based on artificial intelligence, characterized in that: The system comprises: Data acquisition module: obtains multi-source data through multiple channels, builds user portraits based on the obtained multi-source data, and trains user data through deep learning models to identify user preferences, habits and potential needs; Real-time monitoring module: monitors server load, game popularity and user distribution in real time. Based on historical data, it uses time series analysis to predict game peak hours and dynamically adjusts resource allocation according to the prediction results. Anomaly detection module: Through natural language processing and image recognition technology, it conducts real-time review of game content, identifies and filters bad information, and uses AI algorithms to detect abnormal behaviors and take defensive measures based on the detection results; Strategy adjustment module: Based on natural language processing and GPT model, it provides 24 / 7 intelligent customer service and adjusts the response strategy according to user emotions.

Images