Deep learning based game user behavior prediction system

Abstract

The application relates to the technical field of user behavior prediction, in particular to a game user behavior prediction system based on deep learning, which discloses a conventional behavior prediction generation module, a conventional behavior prediction execution module and an abnormal behavior rapid determination module. The conventional behavior prediction generation module and the conventional behavior prediction execution module are arranged, various behaviors of a game user are periodically predicted through a long short-term memory model, the behavior prediction sequence is used to ensure that behaviors with high potential abnormal risk of the game user are preferentially predicted, the order and efficiency of the game user behavior prediction are ensured, the abnormal behavior rapid determination module is arranged, the prediction order of behaviors is dynamically adjusted after abnormal behaviors of the game user are found, the abnormal correlation between the abnormal behaviors and the remaining behaviors is deeply analyzed, and all abnormal behaviors of the game user can be quickly locked in a short time.

Description

Technical Field

[0001] This invention relates to the field of user behavior prediction technology, and more specifically, to a deep learning-based game user behavior prediction system. Background Technology

[0002] With the rapid development of internet technology, online games have become an important form of leisure and entertainment. The number of game users is enormous and their behaviors are diverse. Effectively managing and predicting these user behaviors is crucial for improving the gaming experience, ensuring game security, and optimizing operational strategies. Traditional user behavior prediction methods, such as statistical analysis models or simple machine learning algorithms, often struggle to capture the complexity and dynamism of user behavior, especially when dealing with time-series data and potentially anomalous behaviors.

[0003] In recent years, deep learning technology has demonstrated enormous potential in the field of game user behavior prediction due to its powerful feature extraction and pattern recognition capabilities. In particular, the Long Short-Term Memory (LSTM) model, as a special type of recurrent neural network (RNN), can effectively handle long-term dependencies in time-series data and is suitable for periodic prediction of user behavior. However, existing deep learning-based prediction systems mostly focus on improving prediction accuracy, while improvements are still needed in the orderliness, efficiency, and rapid identification of abnormal behavior.

[0004] In practical applications, game user behavior often exhibits certain periodic patterns, but it also contains potential abnormal behaviors such as cheating and fraud. These abnormal behaviors not only affect the fairness of the game but may also damage the reputation and economic interests of game operators. Therefore, how to achieve efficient prediction and rapid identification of potential abnormal behaviors while ensuring prediction accuracy has become a key problem that urgently needs to be solved in the field of game user behavior prediction.

[0005] To address the aforementioned issues, while some existing technologies attempt to identify abnormal behavior by introducing rule engines or anomaly detection algorithms, these methods often rely on manually set rules and thresholds, lacking adaptability and flexibility. Furthermore, when processing large-scale user behavior data, these methods may suffer from low computational efficiency and high false positive rates.

[0006] Therefore, there is an urgent need for a new predictive system that can comprehensively consider the periodicity of user behavior, potential abnormal risks, and the orderliness and efficiency of prediction. This system should be able to utilize deep learning technology, especially long short-term memory models, to accurately predict various behaviors of game users, and through reasonable module design and algorithm optimization, achieve rapid identification and locking of abnormal behaviors, thereby providing game operators with strong data support and decision-making basis.

[0007] Based on the above background, this invention proposes a game user behavior prediction system based on deep learning. Summary of the Invention

[0008] In view of the shortcomings of existing technologies, the purpose of this invention is to provide a game user behavior prediction system based on deep learning.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A deep learning-based game user behavior prediction system includes a regular behavior prediction sequencing module, a regular behavior prediction execution module, and an abnormal behavior rapid identification module.

[0011] The conventional behavior prediction sequence generation module is used to periodically generate behavior prediction sequences for each game user, and predict each type of game behavior of the game user in turn based on the behavior prediction sequences.

[0012] After predicting a game behavior for each pair of game users, the regular behavior prediction execution module obtains the behavior prediction value for that game behavior. Based on the comparison between the behavior prediction value and the behavior prediction threshold value, it determines whether the game behavior is abnormal.

[0013] When abnormal game behavior occurs, the abnormal behavior rapid determination module marks all game behaviors following the abnormal game behavior in the behavior prediction sequence as abnormal unsolved behaviors, obtains the abnormal synchronization value of each abnormal unsolved behavior, sorts all abnormal unsolved behaviors in ascending order according to the value of the abnormal synchronization value, generates the abnormal rapid determination sequence of the game user, and predicts the abnormal unsolved behaviors of the game user in sequence based on the abnormal rapid determination sequence.

[0014] Furthermore, the process of generating the game user's behavior prediction sequence is as follows: obtain the potential anomalies of a game user for various game behaviors, sort the various game behaviors in descending order of the potential anomaly values, and then generate the game user's behavior prediction sequence.

[0015] Furthermore, the process for obtaining the potential anomaly value of game behavior is as follows: Obtain the predicted value of a game behavior in the previous j periods. Sort all the predicted values ​​of behavior in chronological order according to the period. Compare the next adjacent predicted value of behavior with the previous predicted value of behavior. When the next predicted value of behavior is greater than the previous predicted value of behavior, increase the number of aberrations by one. When the next predicted value of behavior is less than or equal to the previous predicted value of behavior, do not take any action. Mark the number of aberrations as Lvm. Calculate the difference between two adjacent predicted values ​​of behavior after sorting and take the absolute value to obtain the behavior swing value. Sum all the behavior swing values ​​and take the average to obtain the behavior prediction swing value Sew. Sum all the behavior prediction values ​​and take the average to obtain the behavior prediction mean Mds. Use the formula Bzx = Lvm * (Sew * y1 + Mds * y2) to obtain the potential anomaly value Bzx of this game behavior, where y1 is the prediction swing coefficient and y2 is the behavior prediction coefficient.

[0016] Furthermore, after predicting a game behavior for each pair of game users, the predicted value of that game behavior is obtained. Specifically, the game behavior feature set of the game user for a certain game behavior is obtained, the game behavior prediction model of the game user for that game behavior is obtained, the game behavior feature set is used as the input data of the game behavior prediction model, and the predicted value of that game behavior is output.

[0017] Furthermore, the process of obtaining the game behavior feature set for a game user's specific game behavior is as follows: obtain the game behavior features of the game user for the same game behavior in the previous x periods, and assemble all game behavior features into a game behavior feature set in the form of a time series dataset.

[0018] Furthermore, the process of acquiring game behavior features over a period of time is as follows: collect all game behavior data of game users for a certain game behavior over a period of time, extract features from all game behavior data, and obtain game behavior features.

[0019] Furthermore, the abnormal synchronization value of the unresolved abnormal behavior is obtained as follows: determine an unresolved abnormal behavior, obtain the historical behavior analysis value of the unresolved abnormal behavior, synchronously obtain the historical behavior analysis value of the abnormal game behavior, calculate the difference between the historical behavior analysis value of the unresolved abnormal behavior and the historical behavior analysis value of the abnormal game behavior, and take the absolute value to obtain the abnormal synchronization value of the unresolved abnormal behavior.

[0020] Furthermore, the historical behavior analysis value acquisition process is as follows: Obtain the behavior prediction value of a game behavior in the previous d periods. Set a behavior prediction high value and a behavior prediction low value. When the behavior prediction value is greater than or equal to the behavior prediction high value, increment the prediction outlier count by one and mark it as Phm. When the behavior prediction value is less than or equal to the behavior prediction low value, increment the prediction ideal count by one and mark it as Xus. When the behavior prediction value is between the behavior prediction high value and the behavior prediction low value, mark the behavior prediction value as the normal prediction value. Summate all the normal prediction values ​​and take the average to obtain the average normal prediction value Fzw. Use the formula... The historical behavior analysis value Ldt for this type of game behavior is obtained, where a1 is the prediction supervariant coefficient, a2 is the prediction ideal coefficient, and a3 is the conventional prediction coefficient.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. Set up a regular behavior prediction generation module and a regular behavior prediction execution module. Use a long short-term memory model to periodically predict various behaviors of game users. Use behavior prediction sequence to ensure that behaviors with high potential abnormal risk of game users are predicted first, thus ensuring the orderliness and efficiency of game user behavior prediction.

[0023] 2. Set up an abnormal behavior rapid identification module. After detecting abnormal behavior of game users, dynamically adjust the prediction order of behaviors and deeply analyze the abnormal correlation between abnormal behaviors and other behaviors to ensure that all abnormal behaviors of game users can be quickly identified in a short period of time. Attached Figure Description

[0024] Figure 1 This is a system operation flowchart for a deep learning-based game user behavior prediction system.

[0025] Figure 2 This is a system module diagram of a deep learning-based game user behavior prediction system.

[0026] Figure 3 A flowchart for generating sequences to quickly identify anomalies. Detailed Implementation

[0027] Reference Figures 1-3 A deep learning-based game user behavior prediction system includes a regular behavior prediction generation module, a regular behavior prediction execution module, and an abnormal behavior rapid identification module.

[0028] Regular Behavior Prediction Sequence Module: Periodically generates behavior prediction sequences for each game user (one behavior prediction sequence for each game user), and predicts each type of game behavior of the game user in turn based on the behavior prediction sequence (the types of game behavior include interactive behavior, consumption behavior, collection behavior, leisure behavior, etc.).

[0029] The process of generating a prediction sequence of game user behavior is as follows: obtain the potential anomalies of a game user for various game behaviors, sort the various game behaviors in descending order of the potential anomaly values, and then generate the prediction sequence of the game user's behavior.

[0030] The process for obtaining the potential anomaly value of game behavior is as follows: Obtain the predicted value of a game behavior in the previous j periods. Sort all the predicted values ​​of behavior in chronological order according to the period. Compare the next adjacent predicted value of behavior with the previous predicted value. When the next predicted value of behavior is greater than the previous predicted value, increase the number of aberrations by one. When the next predicted value of behavior is less than or equal to the previous predicted value of behavior, do not take any action. Mark the number of aberrations as Lvm. Calculate the difference between two adjacent predicted values ​​of behavior after sorting and take the absolute value to obtain the behavior swing value. Sum all the behavior swing values ​​and take the average to obtain the behavior prediction swing value Sew. Sum all the behavior prediction values ​​and take the average to obtain the behavior prediction mean Mds. Use the formula Bzx = Lvm * (Sew * y1 + Mds * y2) to obtain the potential anomaly value Bzx of this game behavior, where y1 is the prediction swing coefficient and y2 is the behavior prediction coefficient. The value of y1 is 0.84 and the value of y2 is 0.43.

[0031] Regular behavior prediction execution module: After predicting a game behavior of each pair of game users, obtain the behavior prediction value of that game behavior, set the behavior prediction threshold (the behavior prediction threshold is the threshold value), when the behavior prediction value of the game behavior is greater than the behavior prediction threshold value, the game behavior is marked as abnormal game behavior, when the behavior prediction value of the game behavior is less than or equal to the behavior prediction threshold value, continue to predict the next game behavior based on the behavior prediction sequence.

[0032] After predicting a game behavior of each pair of game users, obtain the predicted value of that game behavior. Specifically, obtain the game behavior feature set of the game user for a certain game behavior, obtain the game behavior prediction model of the game user for that game behavior, use the game behavior feature set as the input data of the game behavior prediction model, and output the predicted value of that game behavior.

[0033] The process of obtaining the game behavior feature set of a game user for a certain game behavior is as follows: obtain the game behavior features of the game user for the same game behavior in the previous x periods (each period corresponds to one game behavior feature), and assemble all game behavior features into a game behavior feature set in the form of a time series dataset.

[0034] The process of obtaining game behavior features for a certain game behavior over a period of time is as follows: Collect all game behavior data of game users for a certain game behavior over a period of time, extract features from all game behavior data, and obtain game behavior features.

[0035] Different game behaviors of the same game user correspond to different game behavior prediction models. For example, the interaction behavior and consumption behavior of game user A correspond to two different game behavior prediction models. The same game behavior of different game users also corresponds to different game behavior prediction models. For example, the interaction behavior of game user A and the interaction behavior of game user B correspond to two different game behavior prediction models. This embodiment takes the interaction behavior of game user A as an example and discloses the construction process of the game behavior prediction model for the interaction behavior of game user A: collect multiple game behavior feature sets of game user A for interaction behavior, and construct LSTM. The model uses the game behavior feature set as the training data for the LSTM model. Each training data point is assigned a behavior prediction value, which ranges from 1.1 to 3.9. The closer the behavior prediction value is to 3.9, the more abnormal the interaction behavior of game user A is; the closer the behavior prediction value is to 1.1, the more normal the interaction behavior of game user A is. The training data is divided into a training set, a validation set, and a validation set in a 5:1:1 ratio. The training set, validation set, and validation set are then trained. After training, a game behavior prediction model for the interaction behavior of game user A is constructed.

[0036] The system sets up a regular behavior prediction generation module and a regular behavior prediction execution module. It uses a long short-term memory model to periodically predict various behaviors of game users. The behavior prediction sequence ensures that behaviors with high potential abnormal risks of game users are predicted first, thus ensuring the orderliness and efficiency of game user behavior prediction.

[0037] Abnormal Behavior Rapid Identification Module: When abnormal game behavior occurs, all game behaviors following the abnormal game behavior in the behavior prediction sequence are marked as abnormal unsolved behaviors. The abnormal synchronization value of each abnormal unsolved behavior is obtained. All abnormal unsolved behaviors are sorted in ascending order according to the value of the abnormal synchronization value to generate the abnormal rapid identification sequence for the game user. Based on the abnormal rapid identification sequence, the abnormal unsolved behaviors of the game user are predicted sequentially.

[0038] The abnormal synchronization value of an abnormal unresolved behavior is obtained as follows: Determine an abnormal unresolved behavior, obtain the historical behavior analysis value of the abnormal unresolved behavior, synchronously obtain the historical behavior analysis value of the abnormal game behavior, calculate the difference between the historical behavior analysis value of the abnormal unresolved behavior and the historical behavior analysis value of the abnormal game behavior, and take the absolute value to obtain the abnormal synchronization value of the abnormal unresolved behavior.

[0039] The historical behavior analysis value acquisition process is as follows: Obtain the behavior prediction value of a game behavior in the previous d periods. Set the behavior prediction high value and behavior prediction low value (the behavior prediction high value is greater than the behavior prediction low value, and both the behavior prediction high value and behavior prediction low value are system preset indices). When the behavior prediction value is greater than or equal to the behavior prediction high value, increment the prediction outlier count by one and mark the prediction outlier count as Phm. When the behavior prediction value is less than or equal to the behavior prediction low value, increment the prediction ideal count by one and mark the prediction ideal count as Xus. When the behavior prediction value is between the behavior prediction high value and the behavior prediction low value, mark the behavior prediction value as the normal prediction value. Summate all the normal prediction values ​​and take the average to obtain the average normal prediction value Fzw. Use the formula... The historical behavior analysis value Ldt for this type of game behavior is obtained, where a1 is the prediction supervariant coefficient, a2 is the prediction ideal coefficient, and a3 is the conventional prediction coefficient. The value of a1 is 1.03, the value of a2 is 1.01, and the value of a3 is 0.81.

[0040] The module for quickly identifying abnormal behavior dynamically adjusts the prediction order of behaviors after detecting abnormal behavior of game users, and deeply analyzes the abnormal correlation between abnormal behavior and other behaviors to ensure that all abnormal behaviors of game users can be quickly identified in a short period of time.

[0041] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0042] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0043] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0044] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0045] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0046] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0047] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A deep learning based game user behavior prediction system, characterized by, The method comprises a conventional behavior prediction sequence module, a conventional behavior prediction execution module and an abnormal behavior rapid determination module. The conventional behavior prediction sequence module is configured to periodically generate a behavior prediction sequence for each game user, and sequentially predict each game behavior of the game user based on the behavior prediction sequence. The behavior prediction sequence of the game user is generated by obtaining potential difference values of the game user facing various game behaviors, sorting the various game behaviors according to the values of the potential difference values from large to small, and then generating the behavior prediction sequence of the game user. The potential heterogeneity value acquisition process of the game behavior is as follows: the behavior prediction value of a kind of game behavior in the previous j period is acquired, all the behavior prediction values are sequentially sorted according to the order of the period, the adjacent next behavior prediction value and the previous behavior prediction value after sorting are compared, when the next behavior prediction value is greater than the previous behavior prediction value, the abnormal aggravation times are increased once, when the next behavior prediction value is less than or equal to the previous behavior prediction value, no processing is performed, the abnormal aggravation times are marked as Lvm, the difference value of the adjacent two behavior prediction values after sorting is calculated and the absolute value is taken to obtain the behavior swing value, the sum of all the behavior swing values is processed and the mean value is taken to obtain the behavior prediction swing value Sew, the sum of all the behavior prediction values is processed and the mean value is taken to obtain the behavior prediction mean value Mds, and the formula The potential heterogeneity value Bzx of the game behavior is obtained, wherein y1 is a prediction swing coefficient, and y2 is a behavior prediction coefficient. The conventional behavior prediction execution module obtains a behavior prediction value of each game behavior of each pair of game users after predicting the game behavior, and determines whether the game behavior is an abnormal game behavior based on a comparison result of the behavior prediction value and a behavior prediction threshold value. The abnormal behavior rapid determination module marks all game behaviors after the abnormal game behavior in the behavior prediction sequence as abnormal behaviors to be solved when the abnormal game behavior occurs, obtains abnormal synchronization values of the abnormal behaviors to be solved, sorts all the abnormal behaviors to be solved according to the values of the abnormal synchronization values from small to large, generates an abnormal rapid determination sequence of the game user, and sequentially predicts the abnormal behaviors to be solved of the game user based on the abnormal rapid determination sequence. The abnormal synchronization value of the abnormal behavior to be solved is obtained by determining an abnormal behavior to be solved, obtaining a historical behavior analysis value of the abnormal behavior to be solved, synchronously obtaining a historical behavior analysis value of the abnormal game behavior, performing difference calculation on the historical behavior analysis value of the abnormal behavior to be solved and the historical behavior analysis value of the abnormal game behavior and taking an absolute value to obtain the abnormal synchronization value of the abnormal behavior to be solved. The historical behavior analysis value obtaining process is as follows: obtaining a behavior prediction value of a game behavior in previous d periods, setting a behavior prediction high value and a behavior prediction low value, when the behavior prediction value is greater than or equal to the behavior prediction high value, increasing a prediction super-strange number by one, marking the prediction super-strange number as Phm, when the behavior prediction value is less than or equal to the behavior prediction low value, increasing a prediction ideal number by one, marking the prediction ideal number as Xus, when the behavior prediction value is between the behavior prediction high value and the behavior prediction low value, marking the behavior prediction value as a regular prediction value, summing all the regular prediction values and taking an average to obtain an average regular prediction value Fzw, and obtaining a historical behavior analysis value Ldt of the game behavior by using a formula wherein a1 is a prediction super-strange coefficient, a2 is a prediction ideal coefficient, and a3 is a regular prediction coefficient. 2.The deep learning based game user behavior prediction system of claim 1, wherein, After predicting each game behavior of each pair of game users, the behavior prediction value of the game behavior is obtained, specifically: obtaining a game behavior feature set of the game user for a game behavior, obtaining a game behavior prediction model of the game user for the game behavior, taking the game behavior feature set as input data of the game behavior prediction model, and outputting to obtain the behavior prediction value of the game behavior. 3.The deep learning based game user behavior prediction system of claim 2, wherein, The game behavior feature set of the game user for a game behavior is obtained by obtaining game behavior features of the game user for the same game behavior in the previous x periods, and combining all the game behavior features into a game behavior feature set in the form of a time series data set. 4.The deep learning based game user behavior prediction system of claim 3, wherein, The game behavior feature of a game behavior in a period is obtained by collecting all game behavior data of the game user for a game behavior in a period, and performing feature extraction on all the game behavior data to obtain the game behavior feature.

Images