Game completion guide method and device, computer device, and storage medium

Abstract

The application relates to a game pass guiding method and device, computer equipment and a storage medium. The method comprises the following steps: in response to monitoring that a target account fails in a target level challenge, obtaining current game data of the target account; obtaining a pass benchmark model of the target level, the pass benchmark model being established based on pass data of a game account that successfully passes the target level; identifying at least one short board information of the target account relative to the pass benchmark model; wherein the short board information is used to indicate a game data weakness of the target account; inputting the short board information into a pre-trained guiding model, generating guiding information for the target account through the guiding model; and providing the guiding information to the target account through an in-game user interface. The method can provide game pass guidance highly related to the game data of the account.

Description

Technical Field

[0001] This application relates to the field of game development technology, and in particular to a game level completion guidance method, device, computer equipment, and storage medium. Background Technology

[0002] In massively multiplayer online role-playing games (MMORPGs), player churn due to frustration with level challenges is one of the core issues affecting game operation. Currently, common solutions for players getting stuck on levels mainly include: guiding players to consult community guides outside the game, providing static general hints within the game, or offering options to reduce the difficulty of levels after players have failed multiple times.

[0003] However, these existing solutions share a common major technical problem: a lack of clear gameplay guidance that is highly relevant to the account's own situation during the game. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, computer equipment, and storage medium that can provide game completion guidance that is highly related to the account's own game data, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for guiding players through a game, the method comprising:

[0006] In response to the detection that a target account has failed to complete a target level challenge, the current game data of the target account is obtained; Obtain the benchmark model for the target level, which is built based on the completion data of game accounts that have successfully completed the target level; Identify at least one weakness of the target account relative to the benchmark model; wherein the weakness information is used to indicate the game data weaknesses of the target account; The shortcomings information is input into a pre-trained guidance model, which then generates guidance information for the target account. Provide guidance information to the target account through the in-game user interface.

[0007] In one embodiment, a level completion benchmark model is established based on the level completion data of game accounts that have successfully completed the target level, including: Cluster analysis was performed on the clearance data to identify at least one mainstream clearance strategy for the target level; For each mainstream level-clearing strategy, extract the key level-clearing features of the mainstream level-clearing strategy; Based on key customs clearance features, determine the benchmark sub-models corresponding to mainstream customs clearance strategies; Based on the benchmark sub-models corresponding to multiple mainstream level-clearing strategies, a level-clearing benchmark model corresponding to the target level is established.

[0008] In one embodiment, the key clearance features include at least one of the following: The information includes team character level, core combat attributes, equipment rating, class lineup composition, and the carrying information of specific key items; among them, if the key clearance feature is a numerical feature, then the numerical feature is the mean or median.

[0009] In one embodiment, cluster analysis is performed on the customs clearance data, including: A pre-defined clustering algorithm is used to determine the intra-cluster dispersion variation information under different cluster size values; Based on the information on the change in intra-cluster dispersion, the target number of clusters is determined. The target number of clusters is a value that makes the intra-cluster dispersion relatively flat as the target number of clusters increases. The target number of clusters is used to determine the number of clusters.

[0010] In one embodiment, identifying at least one deficiency information of the target account relative to the clearance benchmark model includes: Compare the target account's current game data with the statistical values ​​corresponding to the key clearance features contained in the clearance benchmark model; The character gap level of the target account is determined based on the quantile of the target account's current game data in the data distribution of the completion data; Based on the role gap level, generate gap information including the type of gap, the gap value of the gap, and the priority of the pass feature.

[0011] In one embodiment, determining the role gap level of the target account includes: If the statistical value corresponding to the key clearance characteristics of the target account is lower than the first threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a weakness. If the statistical value corresponding to the key clearance characteristics of the target account is lower than the second threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a significant gap. The second threshold is lower than the first threshold.

[0012] In one embodiment, the guiding model is a large language model; Before inputting the weakness information into the pre-trained guided model, the following steps are also included: Obtain a predefined prompt template that includes virtual wizard role settings and output format requirements; The information about shortcomings is combined with the prompt word template to obtain the input information for the guidance model; and the input information is used as the input for the guidance model.

[0013] In one embodiment, the prompt word template is used to instruct the guidance model to generate dialogue text that conforms to the virtual guide role settings.

[0014] In one embodiment, the customs clearance data is anonymized log data, which is used to prevent the data from being associated with the real account identity; After establishing a benchmark model for clearing the target level, the following is also included: Discard the original customs clearance data.

[0015] In one embodiment, in response to detecting that a target account has failed a target level challenge, the following includes: If the target account is detected to meet the preset failure conditions, the target level challenge will fail.

[0016] In one embodiment, the method of providing guidance information to the target account includes at least one of dialogue text containing guidance information and voice broadcast of guidance information.

[0017] In one embodiment, the method further includes: Periodically or in response to game version updates, update the completion data of the account group that has successfully completed the target level; The customs clearance benchmark model is updated based on the updated customs clearance data.

[0018] Secondly, this application also provides a game completion guidance device, the device comprising: The failure data acquisition module is used to acquire the current game data of the target account in response to the detection that the target account has failed in the target level challenge; The benchmark model acquisition module is used to acquire the benchmark model of the target level. The benchmark model is built based on the level completion data of the game account that has successfully completed the target level. The weakness identification module is used to identify at least one weakness of the target account relative to the benchmark model; wherein the weakness information is used to indicate the game data weakness of the target account; The guidance generation module is used to input the weakness information into the pre-trained guidance model, and through the guidance model, generate guidance information for the target account; The guide module is used to provide guidance information to the target account through the in-game user interface.

[0019] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0020] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0021] The game completion guidance method, device, computer equipment, and storage medium provided in this application respond to the failure of a target account in a challenge and acquire its current game data. This ensures that the diagnostic analysis is based on the real-time status of the target account in the latest failure scenario, guaranteeing the accuracy and timeliness of the basic data for analysis. By acquiring the completion data of game accounts that have successfully completed the target level, a completion benchmark model is established, providing a quantifiable and objective reference standard for the guidance process based on real success experience, and providing a data foundation for personalized guidance. On this basis, shortcomings relative to the completion benchmark model are identified. By comparing the target account's own real-time game data with the success benchmark in multiple dimensions, the unique and specific data-level weaknesses of the target account can be accurately located, so that the guidance information suggestions of the guidance model are fundamentally based on the specific situation of the target account itself. By inputting information about shortcomings into the guidance model to generate guidance information, and utilizing intelligent technologies such as large-scale language models, technical shortcomings can be transformed into natural, friendly, and actionable language suggestions. Highly relevant data analysis results can be effectively conveyed to players, and guidance information is provided through the game user interface. The generated personalized guidance is seamlessly integrated into the game world and presented through virtual guides using native interactive methods such as dialogue and voice, ensuring immersion and a consistent experience throughout the guidance process. Therefore, this application systematically avoids the technical defects of generalized guidance content, detachment from the player's specific context, and disruption of the game experience by establishing benchmarks through data modeling, capturing failure data in real time, accurately identifying shortcomings through comparison, intelligently converting and generating suggestions, and delivering guidance in an immersive manner. It can provide precise guidance to stuck accounts (players) that is highly relevant to their own character development, equipment configuration, team composition, and other game data weaknesses. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a game completion guidance method provided in an embodiment of this application; Figure 2 This application provides a schematic flowchart illustrating the steps for establishing a pass benchmark model for a target level. Figure 3 This application provides a schematic flowchart illustrating the steps for identifying bottleneck information. Figure 4A schematic diagram of an interface for guiding a target account through a level using a virtual wizard, provided as an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a game level guidance device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] It should be noted that the user information (including but not limited to game accounts, user device information, user personal information, etc.) and data (including but not limited to data used for level completion, game data, analysis data, stored data, display data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0026] In massively multiplayer online role-playing games (MMORPGs), player churn due to level challenges is one of the core issues affecting game operation. Currently, common solutions for players getting stuck include: guiding players to consult community guides outside the game, providing static general hints within the game, or offering options to reduce level difficulty after multiple failures. Specifically, the mainstream in-game solutions for player level-stuck issues fall into three main categories: Community and External Guides: When players encounter setbacks at specific levels, they need to actively interrupt their gameplay and turn to external community forums, strategy websites, or video platforms to search for walkthroughs created and shared by other players. These guides typically present the character setup, combat strategies, or operational techniques required to complete the level in text, images, or video format.

[0027] Static in-game hints: Game designers pre-set fixed hints within levels or related interfaces. Players can trigger these hints through level loading screens, help buttons, or dialogue with non-player characters (NPCs). These hints are mostly general strategic suggestions, such as "be careful to avoid specific boss skills" or "use a certain attribute to attack enemy weaknesses."

[0028] Dynamic Difficulty Adjustment: When the system detects that a player has failed the same level a preset number of times consecutively, it will automatically, or with the player's consent, temporarily or permanently reduce the difficulty of that level. Difficulty adjustments are usually achieved by directly weakening the enemy unit's core combat stats such as health and attack power.

[0029] However, these existing solutions share a common major technical problem: a lack of clear gameplay guidance that is highly relevant to the account's own situation during the game.

[0030] In one exemplary embodiment, Figure 1 This is a flowchart illustrating a game completion guidance method provided in an embodiment of this application, as shown below. Figure 1 As shown, a game completion guidance method is provided. This example illustrates the method's application to a terminal; however, it can also be applied to a server or a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, and tablets. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the method includes the following steps S101 to S105: Wherein: S101. In response to the detection that the target account has failed the target level challenge, obtain the target account's current game data.

[0031] The target account refers to a game account that is currently playing the game and has failed at a specific level. The current game data refers to the target account's real-time information such as character status, equipment configuration, and team composition at the moment of failure (or over a period of time).

[0032] For example, when the terminal detects that a player is using their game account to challenge a certain level (i.e., the target level) and fails, a data capture process can be triggered. The terminal can record all relevant status information of the player's team at this moment to form a failure snapshot data (i.e., the current game data).

[0033] As an example, if a player's character "A" is wiped out while challenging a "boss B", the terminal can record information such as the level, equipment, and current skill combinations of all characters in the team, which can serve as the data basis for subsequent analysis of the failure.

[0034] Optionally, by capturing game data in real time when a challenge is failed, it can be ensured that the information on which the diagnosis is based is up-to-date and fully relevant to the failure situation, avoiding diagnostic bias caused by the use of historical or irrelevant data, and providing timely data input for accurate and contextualized weakness analysis.

[0035] S102. Obtain the benchmark model for the target level. The benchmark model is established based on the level completion data of the game account that has successfully completed the target level.

[0036] Here, "level completion data" refers to a snapshot of a game account's game state when it successfully completes a specific level. "Level completion benchmark model" refers to a standardized data model abstracted from a large (potentially a preset number) of game account data that successfully completes the level, representing one or more effective level completion strategies. This model can serve as a benchmark for judging whether subsequent challengers have met the requirements.

[0037] For example, the terminal running the game, or the server communicating with that terminal, can continuously collect and store key status information when a player (account) successfully completes a target level. This key status information can be processed to construct one or more data models describing how to complete the level, i.e., a level completion benchmark model. This level completion benchmark model can summarize mainstream success experiences, rather than just a single standard strategy answer.

[0038] As an example, the terminal can retrieve data from the server on all player teams (multiple account data) that have successfully cleared a specific dungeon in a game within the past week. This data can include character levels, equipment, team composition, etc., and can be used to establish a baseline model for clearing that dungeon.

[0039] Optionally, by establishing a data-driven benchmark model for clearing levels, vague experience in clearing levels can be transformed into quantifiable data standards, providing an objective and reliable basis for comparison to accurately diagnose the specific shortcomings of individual players and providing a data foundation for achieving highly personalized guidance.

[0040] S103. Identify at least one weakness information of the target account relative to the benchmark model; wherein the weakness information is used to indicate the game data weakness of the target account.

[0041] Among these, "weakness information" refers to specific deficiencies in character development, equipment configuration, and tactical strategies identified by comparing the target account's current game data with a benchmark model for successful gameplay. "Game data weaknesses" can describe the nature of the weakness information, referring to game data items that are significantly below the standards of successful examples in quantitative comparisons.

[0042] For example, the terminal can compare the acquired failure snapshot data with the established benchmark model for that level, item by item or key item by key item. By calculating the gap and determining whether the necessary conditions are met, one or more aspects where the difference between the target account and the success standard data is most significant can be identified.

[0043] As an example, the terminal can compare the team data of character A with the benchmark model of successfully challenging "boss B". It may find that the average level of the team is 5 levels lower than the standard and that it lacks a professional game character with healing ability. Furthermore, based on the above game data weaknesses, the terminal can give the target account relevant tips to improve the success rate of clearing the level.

[0044] Optionally, by performing automated data comparison and weakness identification, the vague sense of frustration a player feels about "I can't beat it" can be transformed into specific, improveable problem points such as "insufficient level and lack of healing." This process can directly map general completion standards to the specific situation of an individual account's completion process, providing a core analytical foundation for providing completion guidance that is highly relevant to the target account's own situation.

[0045] S104. Input the weakness information into the pre-trained guidance model, and generate guidance information for the target account through the guidance model.

[0046] The guidance model refers to a model that can receive data input (which can be structured data) to address the aforementioned shortcomings and output natural language text that meets specific requirements (such as conversational or encouraging language). The guidance information refers to a text and / or speech content output by the guidance model that aims to help and guide the target account to overcome the identified shortcomings.

[0047] For example, the terminal passes bottleneck information generated by the baseline model, which can be described in a structured format (such as lists or key-value pairs), as input parameters to a pre-configured guidance model. This guidance model can then generate targeted and easily understandable suggestion text (guidance information) based on this input bottleneck information.

[0048] As an example, the terminal can input the weakness information "{Weakness 1: Team level is too low, gap: -5 levels; Weakness 2: Team lacks a healer}" into the guidance model, and the guidance model can output: "Your adventure team is commendable for its courage! If everyone can train and improve their levels, they will be more at ease when facing the behemoth. In addition, consider inviting a partner with healing spells to join the next challenge, which will greatly enhance the team's survivability!" Optionally, by using a guided model to transform technically challenging data (such as "-5 level") into warm, natural, and actionable language suggestions, the acceptability of the guided information and the user experience can be improved. This can resolve the confusion and stress that may arise from directly presenting complex numerical comparisons to players, allowing personalized diagnostic results to be effectively communicated in a humanistic way.

[0049] S105. Provide guidance information to the target account through the in-game user interface.

[0050] The in-game user interface can be presented to the user in the form of a virtual guide. This virtual guide can be an intelligent entity or system module within the game with a virtual avatar capable of interacting with the player (such as through dialogue). "Providing" refers to the process of displaying guiding information to the player through the in-game interactive interface.

[0051] For example, the terminal assigns the guidance information generated by the guidance model to a specific agent in the game world (such as a guide NPC, a pet, or a system assistant). This guidance information is then presented to the player of the target account through the agent's interactive interface in the in-game user interface (such as chat bubbles, voice channels, or quest logs).

[0052] As an example, the game's guide sprite can fly to the player's screen and deliver the guidance information generated by S104 through a speech bubble. This guidance information is displayed as an interface and can be accompanied by corresponding voice announcements.

[0053] Alternatively, guidance can be delivered through an in-game virtual wizard, allowing the entire assistance process to be completed within the game world, without requiring players to interrupt the game or leave the current screen to find a guide. This maintains the game's immersion and the continuity of the experience, seamlessly integrating personalized guidance into the player's journey and avoiding disruption to the game's immersion.

[0054] In this embodiment, responding to the target account's failed challenge and acquiring its current game data ensures that the diagnostic analysis is based on the target account's real-time status in the latest failure scenario, guaranteeing the accuracy and timeliness of the basic data for analysis. A benchmark model for clearing a level is established by acquiring the clearing data of game accounts that have successfully cleared the target level. This provides a quantifiable and objective reference standard for the guidance process, based on real-world success experiences, and provides the data foundation for personalized guidance. On this basis, shortcomings relative to the benchmark model are identified. By comparing the target account's real-time game data with the success benchmark in multiple dimensions, the unique and specific data-level weaknesses of the target account can be accurately located. This ensures that the guidance information suggestions of the guidance model are fundamentally based on the specific situation of the target account. Inputting the shortcomings information into the guidance model to generate guidance information, and utilizing intelligent technologies such as large-scale language models, technical shortcomings can be transformed into natural, friendly, and actionable language suggestions. This effectively conveys highly relevant data analysis results to the player. The guidance information is provided through the game user interface, seamlessly integrating the generated personalized guidance into the game world. Presented through virtual guides using native interactive methods such as dialogue and voice, the immersive and consistent experience of the entire guidance process is ensured. Therefore, this application systematically avoids the technical defects of generalized guidance content, detachment from the player's specific context, and damage to the game experience by establishing benchmarks through data modeling, capturing failure data in real time, accurately comparing and identifying shortcomings, intelligently converting and generating suggestions, and delivering immersive completion guidance. It can provide precise guidance to target accounts (players) who are stuck in a level, which is highly related to their own game data weaknesses such as character development, equipment configuration, and lineup combination.

[0055] In one exemplary embodiment, Figure 2 This application provides a schematic flowchart illustrating the steps for establishing a pass benchmark model for a target level, as shown in the embodiments of this application. Figure 2 As shown, step S101, based on the game account's successful completion data benchmark model for the target level, includes: S201. Perform cluster analysis on the clearance data to identify at least one mainstream clearance strategy for the target level. S202. For each mainstream level-clearing strategy, extract the key level-clearing features of the mainstream level-clearing strategy; S203. Based on key customs clearance features, determine the benchmark sub-models corresponding to the mainstream customs clearance strategies; S204. Based on the benchmark sub-models corresponding to multiple mainstream level-clearing strategies, establish a level-clearing benchmark model corresponding to the target level.

[0056] Clustering analysis can be an unsupervised machine learning method used to automatically group similar samples (in this case, data samples from successful players). Mainstream completion strategies refer to common completion methods discovered through clustering, adopted by a significant number of successful players. Key completion features refer to core data dimensions that define and distinguish different completion strategies. A baseline sub-model can be a statistical summary or data profile of the key features of a single mainstream completion strategy.

[0057] For example, the terminal or server running the game can analyze all the collected successful completion data. First, a clustering algorithm can be used to divide these successful cases into several different groups. Each group represents a typical method for completing the level (e.g., high-power overpowered group, attribute-countering technique group, standard configuration for a guaranteed win group). The completion data within each group can be statistically analyzed to extract typical characteristic values ​​for that group (such as average level, common team compositions, etc.). The feature set of each group can constitute a baseline sub-model. The final completion baseline model can be composed of these multiple sub-models.

[0058] As an example, after clustering the successful data of a certain dungeon F, three clusters are obtained: Cluster A (high-level players, any lineup), Cluster B (medium-level players, fixed warrior-mage-priest lineup), and Cluster C (low-level players, dependent on specific summons and control chains). The terminal can build a baseline sub-model for each of these three clusters, and through the three baseline sub-models, construct the benchmark model for clearing the target level.

[0059] In this embodiment, cluster analysis is used to identify multiple mainstream success strategies and establish corresponding sub-models, so that the success benchmark is no longer a single, one-size-fits-all standard. This more realistically reflects the diverse success paths of the player base, allowing subsequent weakness diagnosis to more flexibly and accurately compare failing players with the closest successful strategies, thus providing more valuable and diverse improvement suggestions, further enhancing the personalization and effectiveness of guidance.

[0060] In one exemplary embodiment, the key clearance features include at least one of the following: The information includes team character level, core combat attributes, equipment rating, class lineup composition, and the carrying information of specific key items; among them, if the key clearance feature is a numerical feature, then the numerical feature is the mean or median.

[0061] Among these, core combat attributes refer to numerical values ​​that directly affect combat capabilities, such as attack power, defense, health, and critical hit rate. Equipment rating refers to the overall combat power assessment of a character's equipped gear. Class composition information refers to the proportion of different functional classes (such as tanks, healers, and damage dealers) in the team. Information on the carrying of specific key items refers to whether the character is carrying items that have special effects or are almost essential for this level.

[0062] For example, when constructing a baseline sub-model, the terminal can extract a series of key dimensions from game data of successful cases as key completion features. For key completion features that can be measured numerically (such as level or attack power), the terminal can calculate the central tendency of the key completion feature across all similar successful cases, typically using the average. However, for data that may be affected by extreme values ​​(such as a few players having extremely high scores), the median can be used to better represent the general level. For non-numerical features (such as the presence or absence of a healing class, or whether key equipment is carried), their frequency of occurrence can be recorded or they can be used as necessary conditions.

[0063] As an example, key characteristics for a standard, stable clear strategy might include: an average team level (mean) of 58, a median tank defense of 3200, a team that includes at least one healer in 100% of cases, and 85% of cases having skill J or equipment Z.

[0064] In this embodiment, by clearly defining multi-level and multi-type key clearance features, the clearance benchmark model can comprehensively depict the requirements of a successful strategy from multiple perspectives. This provides rich comparative dimensions for subsequent defect information diagnosis, enabling a more detailed pinpoint whether the player's problem lies in level, attributes, team composition, or item preparation. This generates more comprehensive and specific guidance information, ensuring that the guidance is highly correlated with the account's own data across multiple dimensions.

[0065] In one exemplary embodiment, cluster analysis of customs clearance data includes: A pre-defined clustering algorithm is used to determine the intra-cluster dispersion variation information under different cluster size values; Based on the information on the change in intra-cluster dispersion, the target number of clusters is determined. The target number of clusters is a value that makes the intra-cluster dispersion relatively flat as the target number of clusters increases. The target number of clusters is used to determine the number of clusters.

[0066] The preset clustering algorithm can be a commonly used partitioning clustering method that requires a pre-specified number of clusters, such as the K-Means clustering algorithm. Intra-cluster dispersion refers to the sum or average distance of all sample points within the same cluster to the cluster centroid (center point), used to measure the compactness of samples within a cluster. The variation information refers to the trend of the intra-cluster dispersion value as the K value increases.

[0067] For example, when clustering customs clearance data using the K-Means algorithm, the terminal can try multiple different values ​​of K (e.g., from 2 to 10). For each K value, the terminal can perform clustering and calculate the sum of the intra-cluster dispersion of all clusters. The terminal can then observe the curve of how this sum of dispersion decreases as the K value increases. When the decrease in dispersion due to increasing the K value becomes significantly smaller, it indicates that further increasing the number of clusters no longer contributes much to improving the sample aggregation degree, and the K value at this point can be considered appropriate.

[0068] As an example, the terminal calculates the intra-cluster dispersion for K=2, 3, 4, 5, and 6 respectively. It is found that the dispersion decreases significantly from K=3 to K=4; the decrease is smaller from K=4 to K=5; and the decrease is negligible from K=5 to K=6. Therefore, the terminal can choose K=4 or K=5 as the target cluster size value.

[0069] In this embodiment, the elbow rule, which combines business understanding with data-driven analysis, is used to determine the number of clusters, avoiding model distortion that might result from arbitrarily setting the K value. This ensures that the established level-clearing benchmark model can more realistically and stably reflect the various level-clearing strategy patterns actually existing in the game, laying a more reliable model foundation for generating high-quality, game-realistic, personalized guidance.

[0070] In one exemplary embodiment, Figure 3 A flowchart illustrating the steps for identifying bottleneck information provided in this application embodiment is shown below. Figure 3 As shown, in step S103, identifying at least one deficiency information of the target account relative to the customs clearance benchmark model includes: S301. Compare the current game data of the target account with the statistical values ​​corresponding to the key clearance features contained in the clearance benchmark model; S302. Determine the character gap level of the target account based on the quantile of the target account's current game data in the data distribution of the completion data; S303. Based on the role gap level, generate gap information including the gap type, gap value, and priority of passing features.

[0071] Here, quantiles can be a statistical concept used to represent the value at a specific percentage position in a set of data. Role gap level refers to a qualitative classification of the difference between the target account and the success criteria (e.g., minor deficiency, weak link, significant gap). Weakness type refers to the category to which the gap belongs (e.g., level, attribute, lineup). Weakness gap value refers to a quantified gap value (e.g., -5 level). Feature priority refers to the order in which prompts are given based on the gap level and the importance of the weak link.

[0072] For example, when comparing current game data with the benchmark model, the terminal can compare the absolute difference between the target account's data and the benchmark value, and can also focus on the relative position of that game data among all successfully completed players. For instance, the terminal calculates the percentile ranking of the target account's defense power of 2800 among the defense power data of successful players. If the ranking is in the bottom 15%, it may be determined that a weakness exists. The larger the gap, the higher the priority of the weakness. Finally, the terminal can also organize all identified weakness information into a structured weakness information list.

[0073] As an example, the shortcoming information generated by the terminal after comparison can include a list containing two entries. The first entry is: {"Shortcoming Type": "Level Gap", "Gap Value": -4, "Priority": "High"}, indicating a deficiency in character level, specifically a value below the baseline of 4 units, and this issue is marked as high priority and needs to be addressed first. The second entry is: {"Shortcoming Type": "Missing Character", "Specific Value": "Healer", "Priority": "High"}, indicating that the current team lineup is missing a key healer character, and this issue is also marked as high priority.

[0074] In this embodiment, by introducing the concepts of quantiles and gap levels, the assessment of shortcomings is upgraded from a simple threshold comparison to a relative evaluation based on the overall data distribution. This fully considers the actual distribution of game data (which may not be normally distributed), making the definition of shortcomings more scientific and robust. Combined with priority ranking, this provides a focused basis for the generation of subsequent guidance information, ensuring that guidance prioritizes addressing the most critical account-specific issues that are most disconnected from the distribution of successful completion data.

[0075] In one exemplary embodiment, determining the role gap level of a target account includes: If the statistical value corresponding to the key clearance characteristics of the target account is lower than the first threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a weakness. If the statistical value corresponding to the key clearance characteristics of the target account is lower than the second threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a significant gap. The second threshold is lower than the first threshold.

[0076] The first and second thresholds can both be pre-defined quantile values ​​used to classify the gap levels. For example, the first threshold could be the 20th quantile (i.e., below 80% of successful players), and the second threshold could be the 5th quantile (i.e., below 95% of successful players).

[0077] For example, the terminal can maintain a configuration table to set gap judgment thresholds for different types of completion features. When making comparisons, the terminal can calculate the percentile (or quantile) of the target account's game data among the game data of players who successfully completed the game. This percentile is then compared with a preset threshold. If it is below a second threshold (e.g., 5%), it may indicate that the target account's game data in this area is at an extremely low level among successful players, with a very significant gap. If it is between the second and first thresholds (e.g., 5% to 20%), it can be determined that there is a weakness. If it is above the first threshold, the gap may be considered to be within an acceptable range or not a major problem.

[0078] As an example, for the key feature of tank defense, a first threshold can be set at the 30th percentile, and a second threshold at the 10th percentile. If the target account's defense is at the 15th percentile, it is considered to have a weakness; if it is at the 5th percentile, it is considered to have a significant gap.

[0079] In this embodiment, a clear, configurable, and closely linked-to-the-game ecosystem quantitative assessment mechanism is provided by setting hierarchical, data distribution-based thresholds to determine the level of gaps. This allows the system to accurately distinguish between different levels of failure, such as minor deficiencies, areas requiring improvement, and serious defects, thus providing precise input for generating guidance suggestions that prioritize tasks and are highly matched to the account's own data shortcomings.

[0080] In one exemplary embodiment, the guiding model is a large language model; before inputting the bottleneck information into the pre-trained guiding model, the method further includes: Obtain a predefined prompt template that includes virtual wizard role settings and output format requirements; The information about shortcomings is combined with the prompt word template to obtain the input information for the guidance model; and the input information is used as the input for the guidance model.

[0081] In this context, a large-scale language model can be a deep learning model trained on massive amounts of text data, capable of understanding and generating complex natural language. A cue word template refers to a pre-defined text that guides the language model on how to understand the task, what role it plays, and in what style and format to output the information.

[0082] For example, the terminal loads a pre-written prompt template locally or from a server. This prompt template can define the agent's (virtual guide) personality (e.g., helpful guide), speaking style (e.g., encouraging, conversational), and reserve spaces for inserting gap information. The terminal can then fill the gap information generated in S103 into the designated spaces of the prompt template according to the required format (e.g., a JSON string), combining them into a complete text containing task instructions and contextual information. This combined text is then sent to the API of a large language model for processing.

[0083] As an example, the prompt template is: "You are a lively game assistant named 'Xiao Zhi'. Please generate a dialogue that encourages the player and provides specific improvement suggestions based on the following diagnostic results of a player's failed challenge. Avoid directly mentioning numbers and use conversational language. Diagnostic result: [DIAGNOSIS_PLACEHOLDER]". The terminal replaces the placeholders with the JSON string of the shortcoming information.

[0084] In this embodiment, by using prompt word templates to combine structured weakness information with natural language generation instructions, a large language model is effectively guided to create within a given role and style framework. This method decouples technical data input from human-like language output, allowing the system to flexibly adjust the style of the guiding dialogue without affecting the core diagnostic logic. Ultimately, it generates personalized guiding information that is both tailored to the specific data weaknesses of the account and consistent with the in-game character settings, resulting in a natural and fluent presentation.

[0085] In one exemplary embodiment, the prompt word template is used to instruct the guidance model to generate dialogue text that conforms to the virtual guide role settings.

[0086] For example, the prompt template can contain explicit instructions, prohibiting the listing of numerical values. That is, it prohibits directly and bluntly displaying raw data in the generated text, such as "Your defense is 2800, but the standard is 3500, a difference of 700." Instead, it requires the guiding model to act as a warm, virtual partner, using phrases like "it can be improved further" or "it will be more reliable" to replace specific numerical differences, and using "bring a healing partner" to replace technical terms like "the team lacks a HEALER class." Simultaneously, the prompt template can explicitly prohibit the model from outputting raw numerical comparisons.

[0087] As an example, add the following instructions to the prompt template: "Please use phrases such as 'It is recommended to upgrade your level' or 'You may consider equipping a more durable piece of armor,' and avoid saying things like '5 levels lower' or '700 points lower defense.'" In this embodiment, by generating dialogue text that conforms to the virtual guide's character settings, the prompts can impose strong constraints on the style and format of the output content, ensuring that the final guidance information presented to the player is friendly, easy to accept, and not oppressive. This avoids the potential for secondary frustration caused by directly presenting players with cold, hard numerical discrepancies, transforming the account's own data weaknesses into hopeful improvement suggestions. This enhances the user experience and practical effectiveness of the guidance, allowing highly relevant data-driven instruction to be conveyed in the most effective way.

[0088] In one exemplary embodiment, the clearance data is anonymized log data, which is used to prevent the data from being associated with the real account identity; after establishing a clearance benchmark model for the target checkpoint, the method further includes: discarding the original clearance data.

[0089] Anonymization refers to removing or encrypting any information that can directly or indirectly identify a specific natural person (such as real ID, linked mobile phone number, IP address, etc.) during data collection or transmission, retaining only data related to the game's virtual character. Raw completion data refers to detailed log records of a single player's successful completion of a game, without aggregation or statistical processing.

[0090] For example, when a terminal running the game reports a successful completion log to the server, it can use a virtual character ID or a one-time session ID without attaching any real player information. After receiving this anonymous data, the server can use it for cluster analysis and calculation of the completion benchmark model. Once the completion benchmark model is trained, the original detailed log data used for training will be permanently deleted from the server's temporary storage or memory and will not be retained for a long time.

[0091] As an example, the server receives the log: "{session_id:'xyz789', role_level:60,…}", where the session_id is a one-time event and not bound to the user's real identity. After the baseline model is trained, all original logs containing the session_id are deleted.

[0092] In this embodiment, by implementing compliance measures of collecting data anonymously and discarding it after analysis, the system fully utilizes game data to improve service quality while strictly protecting players' personal privacy and data security. This allows the method of building a personalized guidance model based on massive account data to be implemented compliantly and sustainably, improving the privacy and security of game accounts and ultimately enhancing the gaming experience for all players.

[0093] In one exemplary embodiment, in response to detecting that a target account has failed a target level challenge, the following steps are taken: If the target account is detected to meet the preset failure conditions, the target level challenge will fail.

[0094] Among them, meeting the preset failure conditions can refer to a single complete process from when a player enters the level to when the entire team is wiped out or the mission objective is not achieved and the player exits the level, such as a single failed challenge.

[0095] For example, the terminal can be highly sensitive to the determination of challenge failure, as long as the player fails to meet the completion conditions (such as defeating the final boss) in a single attempt at the target level. Alternatively, the determination of challenge failure can be highly tolerant, for example, it can be considered a failure only if the player fails to pass the target level multiple times in a row. The upper limit of the number of consecutive attempts can be determined by a preset challenge number threshold. Whether this is the player's first attempt or the Nth attempt, the terminal can determine it as meeting the preset failure condition event. Finally, the terminal can also determine the failure condition as met by a challenge failure event manually triggered by the target account (e.g., the target account actively triggers the preset completion guide button, or the target account actively triggers the challenge failure event), and then immediately trigger the subsequent guidance process.

[0096] As an example, if a player attempts level G, floor 10, for the first time and suffers a team wipe during the boss battle, the terminal can detect this failure and initiate the game completion guide process from S102 to S105 mentioned above.

[0097] In this embodiment, by setting the trigger condition to a single failure, the response time from player frustration to receiving help can be greatly shortened. Such an immediate feedback mechanism can intervene when the negative impact on the player's emotions is minimal, providing the most timely guidance that is highly relevant to the current failure situation, effectively preventing the accumulation of frustration, and demonstrating a highly agile guidance service centered on player experience.

[0098] In one exemplary embodiment, providing guidance information to a target account includes at least one of: dialogue text containing the guidance information and voice broadcast of the guidance information.

[0099] For example, in the game's user interface, a speech bubble containing guidance information is displayed, and the guidance information is read aloud by the virtual avatar of the intelligent agent. The speech bubble can be a graphical element commonly used in game UIs (User Interfaces) to display character dialogue, typically accompanied by a speaker's avatar or portrait. The voice reading can refer to the playback of pre-synthesized or real-time generated guidance information via an audio output device.

[0100] For example, after receiving the guidance information text, the terminal can control an in-game intelligent agent character (such as an NPC) to move near the player character or to a specific location on the screen. The guidance text can pop up as a speech bubble above the agent's head. Simultaneously, the terminal can invoke a speech synthesis service to convert the guidance information text into speech, which is then played through the game's sound effects channel, with the agent's lip movements or actions matching accordingly.

[0101] As an example, the agent can walk up to the player with a speech bubble above its head saying, "Hey buddy! Your sails (referring to equipment) look like they need some repairs, or they won't be able to withstand the next storm (referring to the level)." At the same time, the agent's voice can be heard through the speaker.

[0102] Optionally, such as Figure 4 As shown, Figure 4 This is a schematic diagram of an interface for a virtual wizard to guide a target account through a level, as provided in an embodiment of this application.

[0103] In this embodiment, by combining visual (speech bubbles) and auditory (voice broadcast) presentation methods, and relying on the existing virtual guide character within the game for delivery, the expressiveness of the guidance information is enhanced, and the immersive experience is improved. This presentation method, which conforms to the interactive habits of the game world, further strengthens the integration of the guidance process with the game itself, making the professional advice for the account's own problems look and sound like a natural interaction from within the game world.

[0104] In one exemplary embodiment, the method further includes: Periodically or in response to game version updates, update the completion data of the account group that has successfully completed the target level; The customs clearance benchmark model is updated based on the updated customs clearance data.

[0105] Periodicity refers to operations performed at fixed time intervals (such as weekly or monthly). Game version updates refer to the release of new content patches, balance adjustments, or large expansion packs, which may affect the difficulty of levels and gameplay strategies.

[0106] For example, a scheduled task runs in the background of the terminal or server, or the data re-collection and model retraining process is initiated when a change in the game client version number is detected. The system can re-execute steps such as cluster analysis and model building based on player success data within the new time period or under the new version, and replace the old benchmark model with the newly generated benchmark model.

[0107] As an example, after the game releases version 2.0, the monster attributes of a certain dungeon are adjusted. The system can automatically start collecting successful data for that dungeon under version 2.0. A week later, the new data is used to train an updated benchmark model for clearing the dungeon, so as to reflect the mainstream clearing strategies under the current version.

[0108] In this embodiment, by establishing a dynamic model update mechanism, it is ensured that the benchmark model for clearing levels can keep pace with the times and always reflect the best practices of the current game environment and player community. This enables the personalized guidance provided by this application to maintain high relevance and accuracy. Even after multiple version iterations of the game, it can still provide players with effective guidance that is highly matched with the current game ecosystem and their own account data, thus possessing long-term adaptability and vitality.

[0109] In some exemplary implementations, the clearance data may specifically include: character ID, class, level, key combat attributes (such as attack power and defense power), ID and rating of worn equipment, and information on the team's class skills (such as the number of tanks, healers, and DPS classes). During the data cleaning phase, data samples missing key fields can be removed.

[0110] In some exemplary implementations, when the key clearance feature is numerical data such as combat scores that may be unevenly distributed, it is more appropriate to use the median as its statistical value to reduce the influence of extremely high-scoring players on the baseline value.

[0111] In some exemplary implementations, common configurations in key level-clearing features may include, in addition to lineup and key equipment, the carrying status of specific summoned beasts (or pets), depending on the actual business logic of the game.

[0112] In some exemplary implementations, the priority of features in the shortcoming information can be mainly determined by the role gap level, that is, the greater the gap with the successful example data, the higher the corresponding priority.

[0113] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0114] The following describes the game level guidance device provided in the embodiments of this application. The game level guidance device has the same inventive concept as the game level guidance method described above. The solution to the problem provided by the device is similar to the solution described in the above method. Therefore, the specific limitations of one or more game level guidance device embodiments provided below can be referred to the limitations of the game level guidance method above. The game level guidance device described below and the game level guidance method described above can be referred to each other, and will not be repeated here.

[0115] In one exemplary embodiment, Figure 5 This is a schematic diagram of the structure of a game level guidance device provided in an embodiment of this application, as shown below. Figure 5 As shown, the game level guidance device 50 includes: a failure data acquisition module 510, a level-clearing model acquisition module 520, a weakness identification module 530, a guidance generation module 540, and a guidance provision module 550, wherein: The failure data acquisition module 510 is used to acquire the current game data of the target account in response to the detection that the target account has failed in the target level challenge.

[0116] The benchmark model acquisition module 520 is used to acquire the benchmark model of the target level. The benchmark model is built based on the level completion data of the game account that has successfully completed the target level.

[0117] The weakness identification module 530 is used to identify at least one weakness information of the target account relative to the benchmark model for clearing the game; wherein the weakness information is used to indicate the game data weakness of the target account.

[0118] The guidance generation module 540 is used to input the shortcoming information into the pre-trained guidance model, and through the guidance model, generate guidance information for the target account.

[0119] The guide module 550 is used to provide guide information to the target account through the in-game user interface.

[0120] In an exemplary embodiment, the level completion model building module 510 is used to build a level completion benchmark model based on the level completion data of game accounts that have successfully completed the target level, including: performing cluster analysis on the level completion data to identify at least one mainstream level completion strategy for the target level; extracting key level completion features for each mainstream level completion strategy; determining the benchmark sub-model corresponding to the mainstream level completion strategy based on the key level completion features; and building a level completion benchmark model corresponding to the target level based on the benchmark sub-models corresponding to multiple mainstream level completion strategies.

[0121] In an exemplary embodiment, the key clearance feature includes at least one of the following: team character level, core combat attributes, equipment rating, class lineup composition information, and information on the carrying of specific key items; wherein, if the key clearance feature is a numerical feature, the numerical feature is the mean or median.

[0122] In an exemplary embodiment, the customs clearance model building module 510 is used to perform cluster analysis on customs clearance data, including: using a preset clustering algorithm to determine the intra-cluster dispersion change information under different cluster quantity values; and determining the target cluster quantity value based on the intra-cluster dispersion change information, wherein the target cluster quantity value is a value that makes the intra-cluster dispersion relatively flat as the target cluster quantity value increases, and the target cluster quantity value is used to determine the number of clusters.

[0123] In an exemplary embodiment, the weakness identification module 530 is used to identify at least one weakness information of the target account relative to the level-clearing benchmark model, including: comparing the current game data of the target account with the statistical values ​​corresponding to the key level-clearing features contained in the level-clearing benchmark model; determining the role gap level of the target account based on the quantile of the current game data of the target account in the data distribution of the level-clearing data; and generating weakness information including weakness type, weakness gap value and pass feature priority according to the role gap level.

[0124] In an exemplary embodiment, the weakness identification module 530 is used to determine the role gap level of the target account, including: if the quantile of the statistical value corresponding to the key clearance feature of the target account in the clearance data distribution of successfully cleared accounts is lower than a first threshold, it is determined that there is a weakness; if the quantile of the statistical value corresponding to the key clearance feature of the target account in the clearance data distribution of successfully cleared accounts is lower than a second threshold, it is determined that there is a significant gap; wherein, the second threshold is lower than the first threshold.

[0125] In an exemplary embodiment, the guidance model is a large language model; the guidance generation module 540 is used to input the shortcoming information into the pre-trained guidance model, and further includes: obtaining a predefined prompt word template containing virtual guide role settings and output format requirements; combining the shortcoming information with the prompt word template to obtain the input information of the guidance model; and using the input information as the input of the guidance model.

[0126] In one exemplary embodiment, a prompt template is used to instruct the guidance model to generate dialogue text that conforms to the settings of the virtual guide role.

[0127] In an exemplary embodiment, the clearance data is anonymized log data, which is used to prevent the data from being associated with the real account identity; after the clearance model building module 510 builds a clearance benchmark model for the target level, it also includes discarding the original clearance data.

[0128] In an exemplary embodiment, the failure data acquisition module 520 is used to respond to detecting that the target account has failed the target level challenge, including: triggering the target level challenge failure when the target account meets the preset failure conditions.

[0129] In one exemplary embodiment, the guidance providing module 550 provides guidance information to the target account in a manner including at least one of dialogue text containing the guidance information and voice broadcast of the guidance information.

[0130] In an exemplary embodiment, the method implemented by the level completion model building module 510 further includes: periodically or in response to game version updates, updating the level completion data of the account group that has successfully completed the target level; and updating the level completion benchmark model based on the updated level completion data.

[0131] The various modules in the aforementioned game completion guide device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0132] In one exemplary embodiment, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the game completion guidance methods described above.

[0133] In one exemplary embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the game completion guidance methods described above.

[0134] In one exemplary embodiment, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the game completion guidance methods described in the above embodiments.

[0135] Indicatively, such as Figure 6 As shown, Figure 6 This is a schematic diagram of the internal structure of a computer device 600 provided in an embodiment of this application. The computer device 600 can be provided as a server. (Refer to...) Figure 6The computer device 600 includes a processor 602, which further includes one or more processors, and memory resources represented by memory 601 for storing instructions executable by the processor 602, such as a computer program. The computer program stored in memory 601 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 602 is configured to execute instructions to perform the game completion guidance method of any of the above embodiments. The computer device 600 can operate on an operating system stored in memory 601, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0136] The computer device 600 may also include a power supply component 603 configured to perform power management of the computer device 600, a wired or wireless network interface 604 configured to connect the computer device 600 to a network, and an input / output (I / O) interface 605. Wireless operation may be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by a processor, the computer program implements a game progression guidance method. The display unit 607 of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen may be an LCD screen or an e-ink display screen. The input device 606 of the computer device may be a touch layer covering the display screen, buttons, a trackball, or a touchpad located on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

[0137] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0138] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0139] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0140] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for guiding players through game progression, characterized in that, The method includes: In response to the detection that a target account has failed to complete a target level challenge, the current game data of the target account is obtained; Obtain the benchmark model for the target level, which is established based on the level completion data of game accounts that have successfully completed the target level; Identify at least one weakness of the target account relative to the benchmark model; wherein the weakness information is used to indicate the game data weakness of the target account; The aforementioned shortcomings are input into a pre-trained guidance model, which then generates guidance information for the target account. The introductory information is provided to the target account through the in-game user interface.

2. The method according to claim 1, characterized in that, A benchmark model for clearing a level is established based on the clearing data of game accounts that have successfully cleared the target level, including: Cluster analysis is performed on the clearance data to identify at least one mainstream clearance strategy for the target level; For each of the aforementioned mainstream level-clearing strategies, extract the key level-clearing features of the mainstream level-clearing strategy; Based on the key clearance features, a baseline sub-model corresponding to the mainstream clearance strategy is determined; Based on the benchmark sub-models corresponding to the multiple mainstream level-clearing strategies, a level-clearing benchmark model corresponding to the target level is established.

3. The method according to claim 2, characterized in that, The key clearance features include at least one of the following: The team's character level, core combat attributes, equipment rating, class lineup composition information, and information on the carrying of specific key items; wherein, if the key clearance feature is a numerical feature, then the numerical feature is the mean or median.

4. The method according to claim 2, characterized in that, The cluster analysis of the customs clearance data includes: A pre-defined clustering algorithm is used to determine the intra-cluster dispersion variation information under different cluster size values; Based on the intra-cluster dispersion change information, a target cluster quantity value is determined, wherein the target cluster quantity value is a value that makes the intra-cluster dispersion relatively flat as the target cluster quantity value increases, and the target cluster quantity value is used to determine the number of clusters.

5. The method according to claim 1, characterized in that, The identification of at least one deficiency information of the target account relative to the clearance benchmark model includes: The current game data of the target account is compared with the statistical values ​​corresponding to the key clearance features contained in the clearance benchmark model; Based on the quantile of the target account's current game data in the data distribution of the completion data, the character gap level of the target account is determined; Based on the role gap level, generate gap information including the gap type, gap value, and priority of passing features.

6. The method according to claim 5, characterized in that, Determining the role gap level of the target account includes: If the statistical value corresponding to the key clearance feature of the target account is lower than the first threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a weakness. If the statistical value corresponding to the key clearance feature of the target account is lower than the second threshold in the quantile of the clearance data distribution of successfully cleared accounts, it is determined that there is a significant gap. Wherein, the second threshold is lower than the first threshold.

7. The method according to claim 1, characterized in that, The guiding model is a large-scale language model; Before inputting the shortcoming information into the pre-trained guided model, the method further includes: Obtain a predefined prompt template that includes virtual wizard role settings and output format requirements; The shortcoming information is combined with the prompt word template to obtain the input information of the guidance model; and the input information is used as the input of the guidance model.

8. The method according to claim 7, characterized in that, The prompt template is used to instruct the guidance model to generate dialogue text that conforms to the settings of the virtual guide role.

9. The method according to claim 1, characterized in that, The customs clearance data is log data that has been anonymized. The anonymization process is used to prevent the data from being associated with the real account identity. After establishing a benchmark model for passing the target level, the method further includes: Discard the original customs clearance data.

10. The method according to claim 1, characterized in that, The response to detecting that a target account has failed the target level challenge includes: If the target account is detected to meet the preset failure conditions, the target level challenge will fail.

11. The method according to claim 1, characterized in that, The method of providing the guidance information to the target account includes at least one of: dialogue text containing the guidance information and voice broadcast of the guidance information.

12. The method according to claim 1, characterized in that, The method further includes: Periodically or in response to game version updates, update the completion data of the account group that has successfully completed the target level; The customs clearance benchmark model is updated based on the updated customs clearance data.

13. A game level completion guidance device, characterized in that, The device includes: The failure data acquisition module is used to acquire the current game data of the target account in response to the detection that the target account has failed in the target level challenge; The benchmark model acquisition module is used to acquire the benchmark model of the target level, which is established based on the level completion data of the game account that has successfully completed the target level. A weakness identification module is used to identify at least one weakness of the target account relative to the benchmark model; wherein the weakness information is used to indicate the game data weakness of the target account; The guidance generation module is used to input the shortcoming information into a pre-trained guidance model, and through the guidance model, generate guidance information for the target account; The guidance presentation module is used to provide the guidance information to the target account through the in-game user interface.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 12.

Images