Character operation abnormality detection method and terminal

Abstract

An anomaly detection method and terminal for character operations are disclosed. The method monitors the operation data of a virtual character; based on the operation data, it executes a first determination scheme and a second determination scheme. The first determination scheme determines the compatibility between the operation data and the environmental state of the virtual character; the second determination scheme selects a first number of execution paths and determines any missing steps in the operation data based on these paths; a confidence assessment is performed based on the execution results of the first and second determination schemes, and an anomaly determination result is output based on the assessment results. The first determination scheme primarily analyzes the compatibility between the operation data and the environmental state to ensure the rationality of the operation and the consistency of the context; the second determination scheme identifies missing links in the operation by matching execution paths and missing steps. This dual detection mechanism effectively avoids misjudgments caused by a single detection method, comprehensively improving the accuracy of anomaly detection.

Description

Technical Field

[0001] This invention relates to the field of video games, and in particular to a method and terminal for detecting anomalies in character operations. Background Technology

[0002] In modern massively multiplayer online role-playing games (MMORPGs) and other complex action games, players control virtual characters to perform tasks, engage in combat, and explore environments. However, due to the complexity of game quest chains, the diversity of operation modes, and the high flexibility of character controls, abnormal player behavior often occurs. Examples include unexpected skill activations, omissions of key steps in quest chains or disordered operation sequences, and prolonged lingering at specific quest nodes. These abnormal behaviors not only affect the player's gaming experience but can also lead to quest failure, inefficient operations, and even game data anomalies. Accurately identifying abnormal character behavior during game operation is a technical challenge. Existing methods struggle to effectively handle diverse operational scenarios, especially in complex quest chains and varied operational strategies, where traditional detection methods often exhibit high false positive and false negative rates.

[0003] Most common character operation detection methods are based on simple rule matching or single feature analysis. For example, they detect whether skill release is correct by using fixed key combinations or preset operation sequences, or they rely on fixed time thresholds to determine operation stagnation. These methods have several technical shortcomings: First, the detection method is singular and cannot comprehensively consider the operation context and environmental state, leading to easy misjudgment in complex scenarios; second, they lack flexible path matching mechanisms, especially when task chains change and operation strategies are diverse, making it difficult to effectively identify missing steps and operation jumps in the operation path. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and terminal for detecting abnormal character operation, so as to solve the problem that it is difficult to correctly judge when the game character controlled by the player has abnormal behavior.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: An anomaly detection method for character operations includes the following steps: Monitor the operation data of virtual characters; Based on the operation data, a first determination scheme and a second determination scheme are executed. The first determination scheme is used to determine the compatibility between the operation data and the environmental state of the virtual character. The second determination scheme is used to select a first number of execution paths and determine any missing steps in the operation data based on the execution paths. Confidence assessment is performed based on the execution results of the first and second determination schemes, and anomaly determination results are output based on the assessment results.

[0006] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: An anomaly detection terminal for role operation includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps: Monitor the operation data of virtual characters; Based on the operation data, a first determination scheme and a second determination scheme are executed. The first determination scheme is used to determine the compatibility between the operation data and the environmental state of the virtual character. The second determination scheme is used to select a first number of execution paths and determine any missing steps in the operation data based on the execution paths. Confidence assessment is performed based on the execution results of the first and second determination schemes, and anomaly determination results are output based on the assessment results.

[0007] The beneficial effects of this invention are as follows: It provides an anomaly detection method and terminal for role operations. By simultaneously using a first determination scheme and a second determination scheme, the role operation can be comprehensively analyzed from different dimensions. The first determination scheme mainly analyzes the adaptability of operation data and environmental state to ensure the rationality of the operation and the consistency of the context. The second determination scheme identifies missing links in the operation by matching execution paths and missing steps. The dual detection mechanism effectively avoids misjudgments caused by a single detection method, comprehensively improving the accuracy of anomaly detection. Furthermore, through confidence evaluation, the operation matching degree, path selection, and environmental adaptability are comprehensively quantified to generate a confidence value. This avoids making anomaly judgments easily, significantly reduces the false alarm rate, and effectively avoids misjudgments caused by subtle differences in operation or special circumstances. Attached Figure Description

[0008] Figure 1 This is a flowchart of an anomaly detection method for role operations in an embodiment of the present invention; Figure 2 This is a schematic diagram of an anomaly detection terminal for role operation in an embodiment of the present invention; Label Explanation: 1. An anomaly detection terminal for role operation; 2. Memory; 3. Processor. Detailed Implementation

[0009] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0010] Please refer to Figure 1An anomaly detection method for character operations includes the following steps: Monitor the operation data of virtual characters; Based on the operation data, a first determination scheme and a second determination scheme are executed. The first determination scheme is used to determine the compatibility between the operation data and the environmental state of the virtual character. The second determination scheme is used to select a first number of execution paths and determine any missing steps in the operation data based on the execution paths. Confidence assessment is performed based on the execution results of the first and second determination schemes, and anomaly determination results are output based on the assessment results.

[0011] Understandably, monitoring virtual character operation data involves not only capturing players' routine actions (such as movement, attack, and interaction) in real time, but also expanding the data collection dimensions. For example, it involves collecting data on the rhythm of player actions, recording not only the frequency and time intervals of actions, but also analyzing fluctuations in frequency. For instance, during combat, a normal player's action frequency increases as the combat pace accelerates. If a player's action frequency is abnormally stable or fluctuates in a way that doesn't match the combat pace, there may be an operational problem. Regarding the order of actions, it involves not only recording the sequence of actions but also analyzing the frequency of specific action combinations. For example, in role-playing games, certain skill combos have fixed order and combinations of actions. If a player rarely uses or frequently uses these combo combinations incorrectly, it can be identified as abnormal operation. This comprehensive and detailed collection of operation data provides a rich and accurate data foundation for the subsequent accurate detection of abnormal operation patterns.

[0012] Simultaneously, a multi-layered operation behavior model is constructed using deep learning algorithms (such as convolutional neural networks, CNNs). First, a large amount of normal player operation data is preprocessed and features extracted, transforming the data into an image-based data format suitable for CNN processing, for example, mapping operation order and frequency information into a two-dimensional matrix. Then, the CNN model learns the features of normal operation patterns. The model's convolutional layers automatically extract local features from the operation data, while pooling layers reduce data dimensionality and retain key features. After training, the model can perform feature matching and analysis on the input player operation data. When player operation data is input into the model, it outputs a similarity score to the normal operation pattern. If the score is lower than a set threshold, an abnormal operation pattern is identified. Compared to traditional clustering algorithms, this multi-layered operation behavior model can learn the complex features of normal operation patterns more deeply and accurately, improving the accuracy and reliability of abnormal operation detection.

[0013] As can be seen from the above description, the beneficial effects of the present invention are: by monitoring the operation data of virtual characters and executing the first and second determination schemes, comprehensive detection and judgment of operational behaviors are achieved. Specifically: By simultaneously employing both the first and second judgment schemes, a comprehensive analysis of role operations can be conducted from different dimensions. The first judgment scheme primarily analyzes the compatibility between operational data and environmental states, ensuring the rationality of the operation and consistency with the context; the second judgment scheme identifies missing links in the operation by matching execution paths and omitted steps. This dual detection mechanism effectively avoids misjudgments caused by a single detection method, comprehensively improving the accuracy of anomaly detection.

[0014] Because of the variability in character operation data and environmental states, this method flexibly selects appropriate execution paths during detection, combining the fit between virtual task objectives and operation data to effectively handle various operation scenarios. Through dynamic path matching, it can accurately identify operation omissions and errors in complex task chains.

[0015] By comprehensively quantifying the degree of operation matching, path selection and environmental adaptability through confidence assessment, a confidence value is generated, which avoids making anomaly judgments easily, significantly reduces the false alarm rate, and effectively avoids misjudgments caused by subtle differences in operation or special circumstances.

[0016] In some implementations, the second determination scheme is used to select a first number of execution paths and determine the missing steps in the operation data based on the execution paths, specifically including: Based on the operation data, a first number of execution paths are selected. The execution paths are used to achieve the current virtual task objective of the virtual character and the degree of fit with the operation data meets a preset threshold. The timing matching degree between the execution path and the operation data is compared, and a missing step determination result is generated based on the timing matching degree.

[0017] As described above, by selecting the first number of execution paths from a large number of possible paths for matching analysis, the efficiency and accuracy of operation behavior detection are effectively improved. Path selection not only considers the rationality of the operation data but also verifies the fit between the path and the operation data, making the matching between the operation pattern and the expected behavior more scientific and rigorous. Further verification of the rationality of the operation sequence through temporal matching degree can accurately identify anomalies caused by missing or deviated operations. This method effectively reduces the occurrence of false positives and false negatives through multi-layered verification, providing a more stable and accurate solution for anomaly detection in complex task chains.

[0018] For example, in a massively multiplayer online role-playing game (MMORPG), players need to complete a series of quest chains to unlock advanced skills or obtain rare items. To ensure the completeness of the quest flow and the accuracy of the operation steps, a directed graph model is used to detect omissions in these steps, building upon the methods described above. First, the system constructs the game quest flow as a directed graph model, where each node represents a key operation step, and edges represent the sequential relationship (temporal characteristics) between operations. For instance, in a dungeon quest, the quest chain includes the following key steps: "Accept quest" → "Enter dungeon" → "Defeat boss" → "Collect items" → "Return to turn in quest". In this model, each node not only contains a description of the operation action but also is associated with a quest completion status marker and operation temporal information.

[0019] When a player performs an action, the system records the player's action data and path in real time and performs path matching in a directed graph model. For example, if a player completes "entering the instance" and then immediately attempts to "return to complete the quest," skipping the "defeat the boss" and "collect items" steps, the system detects that the action path is interrupted at the "defeat the boss" node and cannot continue, classifying it as an omitted action. To further enhance detection accuracy, the system also incorporates game quest prompts and guidance information for auxiliary judgment. For instance, if the system has already displayed a prompt "Please defeat the boss to complete the quest," and the player still hasn't performed the relevant action, it further confirms that this step has indeed been omitted. This method significantly improves the accuracy of action step omission detection.

[0020] Meanwhile, in complex tasks or puzzle levels, players often spend a considerable amount of time at certain points. To avoid false positives, a basic time threshold is set, and adjustments are made dynamically based on a comprehensive consideration of task complexity and the player's historical action data. For example, in scenarios requiring solving mechanical puzzles, a reasonable dwell time might be longer, while in simple scene transitions, the time threshold is shorter. The system also compares the player's historical dwell time in similar tasks or levels. If the current dwell time significantly exceeds the historical average and differs greatly from the average dwell time of other players in that section, it is judged as excessive dwell time in that section, indicating an anomaly. This multi-dimensional time judgment strategy effectively avoids false positives caused by player strategy adjustments or excessively high game difficulty, further improving the reliability and accuracy of action detection.

[0021] In some implementations, the step of performing confidence assessment based on the execution results of the first and second determination schemes, and outputting anomaly determination results based on the assessment results, specifically includes: Based on the historical operation data of the virtual character, the confidence values ​​of the first determination scheme and the second determination scheme are evaluated respectively, and the anomaly determination result is output based on the comparison result between the confidence value and the preset confidence value.

[0022] As described above, a confidence assessment mechanism comprehensively analyzes the detection results of the first and second judgment schemes, comparing historical operation data of the virtual character with current operation data to generate a confidence value. The calculation of the confidence value considers not only the matching degree of the detection results but also comprehensively evaluates historical operation characteristics and environmental conditions, making anomaly detection more dynamic and flexible. This method avoids misjudgments caused by accidental operational deviations by reasonably setting a pre-set confidence threshold, thus enabling accurate anomaly detection even under significant operational changes, further enhancing the stability and robustness of the detection method.

[0023] In some implementations, the first determination scheme is used to analyze the compatibility between the operation data and the environmental state of the virtual character, specifically including: The interaction sequence, interaction environment, and character state of the virtual character are determined based on the operation data. Analyze the adaptation of the interaction sequence and the interaction environment, and generate a first adaptation result; Analyze the compatibility between the character state and the interaction sequence, and generate a second compatibility result; An adaptation degree is generated regarding the environmental state based on the first adaptation result and the second adaptation result.

[0024] Understandably, in assessing environmental context, besides comparing button combinations, it's also necessary to consider the contextual information of the operation. Taking skill release as an example, it's not only about determining if the button combination is correct, but also considering the character's current state (e.g., whether they are in a state where skills can be released, whether there is a skill cooldown), and the game environment (e.g., whether there are enemies nearby, whether they are in a specific mission area). If the button combination is correct but doesn't fit the contextual conditions, it's also considered an operation error. This avoids misjudgments caused by simply matching button combinations and more accurately identifies player operation errors in actual game situations.

[0025] As described above, the first judgment scheme analyzes the compatibility between operational data and the virtual character's environmental state, effectively solving the problem of misjudgment caused by the mismatch between operational behavior and contextual environment. This method extracts information such as the character's interaction sequence, interaction environment, and character state, comprehensively analyzes the correlation between operational behavior and environmental state, and generates corresponding compatibility values. This approach fully utilizes the contextual features of operational behavior, extending anomaly detection beyond the operational level to include the matching analysis of operations and the environment, effectively improving the accuracy of operational anomaly identification and environmental understanding. By comprehensively judging the rationality of operations based on the compatibility results, the scientific rigor and accuracy of operational anomaly detection are further improved, ensuring the effectiveness of detection results in complex game environments.

[0026] Specifically, the determination steps for the first determination scheme and the second determination scheme are as follows: The operation data is encoded into vector form, and the Euclidean distance or cosine similarity with the vector of the preset target operation is calculated. The corresponding judgment result is generated based on the difference in Euclidean distance, as follows: The first determination scheme includes: Calculate the Euclidean distance between the interaction sequence and the interaction environment and compare it with the first preset target distance to generate the first adaptation result; Calculate the Euclidean distance between the character state and the interaction order and compare it with the second preset target distance to generate a second adaptation result; An adaptation degree is generated regarding the environmental state based on the first adaptation result and the second adaptation result.

[0027] In some implementations, the steps also include: If the anomaly determination result is non-abnormal behavior and the virtual character stays at the current node for more than the preset duration, then a guidance entry is generated by combining the node data and based on historical operation data.

[0028] As described above, when the anomaly assessment indicates non-abnormal behavior, but the virtual character remains at the current node for longer than a preset duration, the system can intelligently generate personalized prompts to help players find the correct operation path more quickly. These prompts are based not only on node data but also on analysis of the player's historical operation habits, generating targeted prompts that effectively reduce stagnation caused by operational confusion or task difficulty. This flexible prompt mechanism avoids directly interfering with the player's operational smoothness while promptly guiding them to overcome operational bottlenecks, significantly improving the game experience and operational guidance effectiveness. Furthermore, the system can dynamically adjust its prompt strategy, optimizing prompt content based on the player's personalized characteristics to further improve the accuracy and usability of the prompts.

[0029] Specifically, the steps for generating the prompt item are as follows: The player profile data management module collects, organizes, and stores player profile data, which includes game history, operation habit data, game style preference data, and interaction data. The profile data is updated in real time or periodically through various stages of the game and is categorized and stored in a structured database to ensure data accuracy and timeliness. The demand detection module monitors player behavior in real time, acquires real-time player operation data, game scene information, and dwell time in specific game segments, and analyzes player behavior according to preset rules and algorithms to determine whether the player needs prompts. In particular, when a player has no operation progress for a long time in a specific level, or when the health value drops rapidly in a battle scene, the module combines historical operation data and similar scene performance to preliminarily determine that the player needs prompts. The decision tree and rule engine fusion module is activated when the demand detection module determines that the player needs a prompt. It first retrieves the player's feature data from the player profile data management module and combines this data with the current game scene information to generate the prompt content; among which: The decision tree algorithm constructs a decision tree based on player characteristics and scene factors. The tree nodes are divided according to the player's game style preferences and scene type, and the prompt strategy is refined layer by layer. The decision tree is trained and optimized through machine learning algorithms to improve the accuracy and effectiveness of prompts. The rules engine defines rules that match the characteristics of the game, serving as a supplementary mechanism to assist the decision tree algorithm. When the decision tree cannot generate clear hints, the rules engine generates hints as a backup solution. The rules engine rules include, but are not limited to, judgment of health thresholds in combat scenarios, suggestions for obtaining items, and hints for puzzle-solving strategies in levels. The rules engine can be dynamically adjusted and expanded based on game updates and player feedback. The prompt content generation and output module generates specific prompt information based on the analysis results of the decision tree and rule engine fusion module. It also extracts relevant clues and content from the level background knowledge database or problem-solving strategy database, and generates prompt text, graphic prompts or voice prompts by combining player characteristics and current operation status. The prompt information is presented through interface display, voice broadcast or NPC dialogue, making the prompt content intuitive, clear and targeted.

[0030] Please refer to Figure 2 An anomaly detection terminal 1 for role operation includes a memory 2, a processor 3, and a computer program stored in the memory 2 and executable on the processor 3. When the processor executes the computer program, it completes the steps in an anomaly detection method for role operation.

[0031] In summary, this invention proposes an anomaly detection method and terminal for character operations. By monitoring the operation data of virtual characters and combining a first judgment scheme and a second judgment scheme, it achieves comprehensive detection and judgment of operational behavior, demonstrating significant practicality and advancement. In complex game environments, player operation patterns are highly diverse and variable, and operational errors and omissions frequently occur in actual operation. Simple single detection mechanisms are insufficient to effectively cope with changing operational scenarios. This invention, by constructing a dual judgment mechanism, not only improves the accuracy of anomaly detection but also effectively avoids misjudgments and omissions caused by individual operational differences.

[0032] This invention achieves comprehensive detection and analysis of operational behavior through the organic combination of a first judgment scheme and a second judgment scheme. The first judgment scheme analyzes the fit between operational data and environmental states to ensure the rationality and contextual consistency of operations, accurately identifying operational errors in the rationality judgment. For example, in skill release operations, it not only verifies whether the key combination is correct but also considers the current state of the virtual character, skill cooldown time, and environmental factors to further evaluate the matching of key operations with the actual game context, effectively avoiding false alarms caused by simple operational deviations. The second judgment scheme mainly focuses on matching operational paths and missing steps. It constructs a task chain using a directed graph model, monitoring and verifying key nodes and operational timing, accurately identifying abnormal behavior caused by operational omissions or path jumps in complex multi-step task chains. This path matching mechanism further enhances the comprehensiveness and accuracy of operational anomaly detection by calculating the fit and timing matching degree between operational data and the expected execution path.

[0033] Furthermore, this invention introduces a confidence assessment mechanism that comprehensively quantifies operation matching degree, path selection, and environmental state adaptability to generate a confidence value. In determining abnormal operational behavior, it does not directly rely on a single indicator, but rather dynamically adjusts the confidence value by calculating the operation difference value and context matching degree, and combining historical operation data, ensuring the scientific rigor and reliability of the determination. By reasonably setting a pre-set confidence threshold, this invention can effectively reduce false positives caused by subtle differences in operation or special circumstances. Especially in scenarios with significant operational changes, the confidence mechanism can flexibly adjust the determination strategy, making anomaly detection more stable and robust. This multi-dimensional assessment method ensures detection accuracy while minimizing false positive and false negative rates.

[0034] Building upon operational behavior monitoring, this invention also effectively avoids prompt failures caused by time misjudgments in complex task chains by dynamically adjusting operation path selection and time thresholds. Particularly in complex puzzle-solving sections, a reasonably long time threshold is set, while the time threshold is shortened in simple scene transitions. Furthermore, by comparing and verifying historical player operation data with the average operation time of other players, it can accurately determine whether a player's dwell time at a certain node is excessive, thereby further improving the accuracy and real-time performance of anomaly detection.

[0035] To further enhance user experience in anomaly detection and player guidance, this invention introduces a prompt and guidance information generation mechanism. When abnormal operations or prolonged pauses are detected, the system dynamically generates targeted prompts by integrating player profile data and real-time operation data. Through the fusion of decision tree algorithms and rule engines, the system can automatically analyze player characteristics and the current game scenario, providing accurate operation suggestions and strategy hints. This prompt mechanism not only responds quickly to operational deviations but also intelligently adjusts prompt strategies in different scenarios. For example, it provides suggestions on skill release and item usage in combat scenarios, and offers problem-solving strategies or key clues in puzzle-solving scenarios. This flexible and diverse prompting method avoids disrupting the player experience with excessive prompts while providing accurate guidance at crucial moments, significantly improving the player's gaming experience and operational smoothness.

[0036] This invention constructs a comprehensive and flexible detection mechanism, organically combining multiple judgment methods and prompting strategies to achieve the dual effect of improving the accuracy of anomaly detection and reducing false positives. Simultaneously, the dynamic update mechanism ensures the system can adapt to game updates and player skill improvements, realizing high practical value and application prospects. Whether in role-playing games or complex quest chain games, this invention can perform real-time and efficient operation monitoring and anomaly detection, and intelligently generate personalized prompts when players encounter difficulties, effectively improving the gaming experience and operational rationality, thus possessing significant promotional value and application significance.

[0037] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for detecting anomalies in character operations, characterized in that: Including the following steps: Monitor the operation data of virtual characters; Based on the operation data, a first determination scheme and a second determination scheme are executed. The first determination scheme is used to determine the compatibility between the operation data and the environmental state of the virtual character. The second determination scheme is used to select a first number of execution paths and determine any missing steps in the operation data based on the execution paths. Confidence assessment is performed based on the execution results of the first and second determination schemes, and anomaly determination results are output based on the assessment results.

2. The method for detecting anomalies in character operations according to claim 1, characterized in that: The second determination scheme is used to select a first number of execution paths and determine the missing steps in the operation data based on the execution paths, specifically including: Based on the operation data, a first number of execution paths are selected. The execution paths are used to achieve the current virtual task objective of the virtual character and the degree of fit with the operation data meets a preset threshold. The timing matching degree between the execution path and the operation data is compared, and a missing step determination result is generated based on the timing matching degree.

3. The method for detecting anomalies in character operations according to claim 1, characterized in that: The step of performing a confidence assessment based on the execution results of the first and second determination schemes, and outputting anomaly determination results based on the assessment results, specifically includes: Based on the historical operation data of the virtual character, the confidence values ​​of the first determination scheme and the second determination scheme are evaluated respectively, and the anomaly determination result is output based on the comparison result between the confidence value and the preset confidence value.

4. The method for detecting anomalies in character operations according to claim 1, characterized in that: The first determination scheme is used to analyze the compatibility between the operation data and the environmental state of the virtual character, specifically including: The interaction sequence, interaction environment, and character state of the virtual character are determined based on the operation data. Analyze the adaptation of the interaction sequence and the interaction environment, and generate a first adaptation result; Analyze the compatibility between the character state and the interaction sequence, and generate a second compatibility result; An adaptation degree is generated regarding the environmental state based on the first adaptation result and the second adaptation result.

5. The method for detecting anomalies in character operations according to claim 1, characterized in that: It also includes the following steps: If the anomaly determination result is non-abnormal behavior and the virtual character stays at the current node for more than the preset duration, then a guidance entry is generated by combining the node data and based on historical operation data.

6. An anomaly detection terminal for role operation, characterized in that: Includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps: Monitor the operation data of virtual characters; Based on the operation data, a first determination scheme and a second determination scheme are executed. The first determination scheme is used to determine the compatibility between the operation data and the environmental state of the virtual character. The second determination scheme is used to select a first number of execution paths and determine any missing steps in the operation data based on the execution paths. Confidence assessment is performed based on the execution results of the first and second determination schemes, and anomaly determination results are output based on the assessment results.

7. The abnormal detection terminal for role operation according to claim 6, characterized in that: The second determination scheme is used to select a first number of execution paths and determine the missing steps in the operation data based on the execution paths, specifically including: Based on the operation data, a first number of execution paths are selected. The execution paths are used to achieve the current virtual task objective of the virtual character and the degree of fit with the operation data meets a preset threshold. The timing matching degree between the execution path and the operation data is compared, and a missing step determination result is generated based on the timing matching degree.

8. The abnormal detection terminal for role operation according to claim 6, characterized in that: The step of performing a confidence assessment based on the execution results of the first and second determination schemes, and outputting anomaly determination results based on the assessment results, specifically includes: Based on the historical operation data of the virtual character, the confidence values ​​of the first determination scheme and the second determination scheme are evaluated respectively, and the anomaly determination result is output based on the comparison result between the confidence value and the preset confidence value.

9. The abnormal detection terminal for role operation according to claim 6, characterized in that: The first determination scheme is used to analyze the compatibility between the operation data and the environmental state of the virtual character, specifically including: The interaction sequence, interaction environment, and character state of the virtual character are determined based on the operation data. Analyze the adaptation of the interaction sequence and the interaction environment, and generate a first adaptation result; Analyze the compatibility between the character state and the interaction sequence, and generate a second compatibility result; An adaptation degree is generated regarding the environmental state based on the first adaptation result and the second adaptation result.

10. An anomaly detection terminal for role operation according to claim 6, characterized in that: It also includes the following steps: If the anomaly determination result is non-abnormal behavior and the virtual character stays at the current node for more than the preset duration, then a guidance entry is generated by combining the node data and based on historical operation data.

Images