A method and system for generating a virtual object attribute adjustment scheme

Abstract

The application discloses a kind of attribute adjustment scheme generation method and system of virtual object, by storing the first interactive behavior data and attribute growth value of virtual object in the first application database, the growth process of virtual object can be dynamically tracked, and attribute growth curve is established according to actual data, to provide data support for subsequent intelligent adjustment;By establishing an evaluation model, the attribute data is associated with interactive behavior performance, the potential and short board of virtual object can be accurately analyzed, and targeted attribute adjustment scheme is provided for users;By integrating the virtual object data shared by the second application database, cross-application general attribute adjustment scheme can be provided for users, repeated cultivation is avoided, cross-application cultivation resources are saved, and the linkage between applications and user experience are enhanced.

Description

Technical Field

[0001] This invention relates to the field of data processing technology for virtual objects, and in particular to a method and system for generating attribute adjustment schemes for virtual objects. Background Technology

[0002] In existing methods for adjusting the attributes of virtual objects (such as pets and characters), players typically rely on fixed in-game instructions or their own experience to develop them. Traditional methods have the following problems: 1. Lack of dynamic analysis and real-time feedback In existing technologies, the attribute adjustment of virtual objects is mainly based on static rules or general suggestions, and cannot be dynamically analyzed based on real-time combat data. For example, when a virtual object suffers continuous high damage in battle, the system cannot respond quickly and provide targeted defensive attribute enhancement schemes. This leads to low player training efficiency and makes it difficult to optimize the combat performance of virtual objects.

[0003] 2. Data isolation and insufficient cross-game compatibility The development data for virtual objects in different games is usually independent and lacks a sharing mechanism. Even if the same virtual object appears in multiple games, its attribute adjustment schemes cannot be shared, causing players to need to repeatedly invest resources and making cross-game collaborative development impossible. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for generating attribute adjustment schemes for virtual objects, which can dynamically analyze combat data and realize cross-game data sharing and personalized recommendations for attribute adjustment schemes.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for generating attribute adjustment schemes for virtual objects includes the following steps: The first interactive behavior data of the virtual object and the growth value of the first attribute data of the virtual object when the level data changes are stored in the first application database in real time. The growth curve of the first attribute data of the virtual object is established based on the growth value of the first attribute data. Based on the first interaction behavior data and the first attribute growth curve of the same type of virtual objects, an evaluation model is established that associates the first attribute data and the first interaction behavior data. In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model; In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

[0006] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A system for generating attribute adjustment schemes for virtual objects 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 implements the various steps of the aforementioned method for generating attribute adjustment schemes for virtual objects.

[0007] The beneficial effects of this invention are as follows: By storing the first interactive behavior data and attribute growth values ​​of virtual objects in the first application database, the growth process of virtual objects can be dynamically tracked, and attribute growth curves can be established based on actual data, thereby providing data support for subsequent intelligent adjustments; by establishing an evaluation model and associating attribute data with interactive behavior performance, the potential and shortcomings of virtual objects can be accurately analyzed, providing users with targeted attribute adjustment solutions; by integrating the virtual object data shared in the second application database, users can be provided with universal attribute adjustment solutions across applications, avoiding redundant training, saving cross-application training resources, and enhancing the linkage between applications and user experience. Attached Figure Description

[0008] Figure 1 This is a flowchart of a method for generating an attribute adjustment scheme for a virtual object according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a virtual object attribute adjustment scheme generation system according to an embodiment of the present invention; Label Explanation: 1. A system for generating attribute adjustment schemes for virtual objects; 2. A memory; 3. A 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 1 This invention provides a method for generating attribute adjustment schemes for virtual objects, comprising the following steps: The first interactive behavior data of the virtual object and the growth value of the first attribute data of the virtual object when the level data changes are stored in the first application database in real time. The growth curve of the first attribute data of the virtual object is established based on the growth value of the first attribute data. Based on the first interaction behavior data and the first attribute growth curve of the same type of virtual objects, an evaluation model is established that associates the first attribute data and the first interaction behavior data. In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model; In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

[0011] As can be seen from the above description, the beneficial effects of the present invention are as follows: By storing the first interactive behavior data and attribute growth values ​​of virtual objects in the first application database, the growth process of virtual objects can be dynamically tracked, and attribute growth curves can be established based on actual data, thereby providing data support for subsequent intelligent adjustments; by establishing an evaluation model and associating attribute data with interactive behavior performance, the potential and shortcomings of virtual objects can be accurately analyzed, providing users with targeted attribute adjustment solutions; by integrating the virtual object data shared in the second application database, users can be provided with universal attribute adjustment solutions across applications, avoiding redundant training, saving training resources across applications, and enhancing the linkage between applications and user experience.

[0012] Furthermore, in response to an operation to view attribute data of the virtual object, the method further includes: Time series analysis is used to analyze the first interaction behavior data and first attribute data of the virtual objects in the first application database within a preset time period to generate an attribute change prediction model. The attribute change prediction model is used to predict the change trend of the first attribute of the virtual object; By combining the corresponding environmental data and corresponding user data of the virtual object, the change trend of the first attribute is adjusted to obtain the change trend of the second attribute; Based on the changing trend of the second attribute, a periodic third attribute adjustment scheme is generated for the virtual object.

[0013] As described above, predicting the attribute change trend of virtual objects through time series analysis can provide users with long-term training plans and avoid resource waste; adjusting the prediction results by combining environmental data and user data can provide personalized attribute adjustment schemes for different users to meet diverse needs; generating periodic attribute adjustment schemes can help users achieve training goals in stages and improve the scientificity and operability of attribute adjustment schemes.

[0014] Furthermore, based on the first interaction behavior data and the first attribute growth curve of virtual objects of the same type, an evaluation model is established that associates the first attribute data and the first interaction behavior data, including: Calculate the first dispersion and first similarity of the first interaction behavior data of virtual objects of the same type, and calculate the second dispersion and second similarity of the first attribute growth curve of virtual objects of the same type. The range of cluster numbers is determined based on the first degree of dispersion, the first similarity, the second degree of dispersion, and the second similarity; The initial clustering data corresponding to each cluster number within the range of cluster numbers are evaluated, and the initial clustering data corresponding to the cluster number with the best evaluation result is taken as the clustering result; If the virtual object has a preset gain scheme at the first level, the clustering result is adjusted according to the preset gain scheme. The preset gain scheme includes adding a preset skill or increasing the value of the attribute data at a preset rate. Extract data features of each type from the clustering results, and establish an evaluation model that associates the first attribute data and the first interaction behavior data by combining the data features of each type.

[0015] As described above, by calculating the dispersion and similarity of interaction behavior data and attribute growth curves, the range of cluster numbers can be scientifically determined, the construction process of the evaluation model can be optimized, and the accuracy and applicability of the model can be improved. By adjusting the clustering results in conjunction with the preset gain scheme, it can be ensured that the model can still provide reasonable evaluation and adjustment suggestions in special cases (such as the addition of new skills or accelerated attribute growth). By extracting data features from the clustering results, a more accurate correlation model between attributes and interaction behavior performance can be established, providing a reliable basis for attribute adjustment.

[0016] Further, in response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model, including: If the interaction with the virtual object is a continuous attack operation that exceeds a preset number of times and a preset damage value, then the evaluation model is used to analyze the reference defense attribute data corresponding to the interaction operation. If the defense attribute value of the virtual object is less than the reference defense attribute data, then based on the difference between the defense attribute value of the virtual object and the reference defense attribute data, virtual items and virtual tasks are selected to improve the defense attribute value of the virtual object, and the virtual items and virtual tasks are used as the first attribute adjustment scheme of the virtual object.

[0017] As described above, when a series of high-damage attacks are detected, the system can quickly analyze the required defensive attributes and provide users with immediate defense enhancement solutions to improve the survivability of virtual objects. It can also select suitable items and tasks based on the difference in defensive attributes and generate adjustment plans through real-time interactive behavior data to ensure the relevance and timeliness of the solutions.

[0018] Furthermore, based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a general second attribute adjustment scheme for the virtual object is established, including: Based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a second attribute growth curve of the shared virtual object is generated. If the first attribute growth curve and the second attribute growth curve have a similarity interval greater than a first threshold, and the similarity between the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval is greater than a second threshold, then a general second attribute adjustment scheme for the virtual object is generated based on the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval.

[0019] As described above, by generating attribute growth curves for shared virtual objects, data from multiple applications can be integrated; by identifying similar intervals and matching interaction behavior data, users can be provided with a universal attribute adjustment scheme across applications; through cross-application strategy sharing, users can learn from the experience of other applications, avoid repeated trial and error, and improve the effectiveness of the generated attribute adjustment scheme.

[0020] Please refer to Figure 2 Another embodiment of the present invention provides a virtual object attribute adjustment scheme generation system, including 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 implements the various steps of the above-described virtual object attribute adjustment scheme generation method.

[0021] The method and system for generating attribute adjustment schemes for virtual objects described above are particularly suitable for generating attribute adjustment schemes to achieve cross-game application data sharing and personalized recommendations. The following detailed implementation methods illustrate these methods: Please refer to Figure 1 Embodiment 1 of the present invention is as follows: A method for generating attribute adjustment schemes for virtual objects includes the following steps: S1. Store the first interaction behavior data of the virtual object in real time in the first application database, and store the growth value of the first attribute data of the virtual object when the level data changes. Establish the first attribute growth curve of the virtual object based on the growth value of the first attribute data.

[0022] In this embodiment, the first application database is the database of the first game application, the virtual object is the game pet object, and the interaction behavior data is the battle data.

[0023] Specifically, the game's client collects various data about virtual objects in real time during combat, including the type and value of damage received, damage output, skill release frequency, attack hit rate, and evasion rate, and stores this data in a database. Simultaneously, it records the increase in various attributes of virtual objects each time they level up, constructing a database of virtual object growth curves.

[0024] Specifically, in the game client, event listeners are deployed using event-driven programming mechanisms in programming languages ​​such as C++ or Java. When a pet enters the battle scene, the listener is triggered, using hook functions (functions that can intercept and process specific events) to capture damage events, skill release events, and so on in real time. For example, in C++, by overriding the handling functions for relevant battle events, data such as the type of damage the pet receives (physical damage, magical damage, etc.) and its value, as well as the damage output, can be obtained. This data is temporarily stored in a local cache (such as using a memory caching library like Memcached) to ensure fast data read and write speeds.

[0025] When a pet level-up event is triggered, the game client connects to the game database via an interface (e.g., using Java's JDBC to connect to a MySQL database) to retrieve the growth values ​​of various attributes at the time of level-up. Using data persistence technology, these values ​​are stored in a constructed pet growth curve database according to a relational database (such as MySQL) table structure. For example, a table is created containing fields such as pet ID, level, and the growth values ​​of each attribute, and the data for each level-up is inserted into this table.

[0026] S2. Based on the first interaction behavior data and the first attribute growth curve of the same type of virtual objects, establish an evaluation model that associates the first attribute data and the first interaction behavior data.

[0027] In this embodiment, machine learning algorithms are used to perform in-depth analysis of the collected data. By comparing the combat data and growth curves of a large number of pets of the same type, a correlation model between attributes and combat performance is established. For example, the analysis found that after a certain type of pet reaches a certain attack power value, its attack hit rate will significantly increase. Based on this model, the pet's innate attributes and growth potential are evaluated.

[0028] S21. Calculate the first dispersion and first similarity of the first interaction behavior data of virtual objects of the same type, and calculate the second dispersion and second similarity of the first attribute growth curve of virtual objects of the same type.

[0029] S22. Determine the range of cluster numbers based on the first dispersion degree, the first similarity, the second dispersion degree, and the second similarity.

[0030] S23. Evaluate the initial clustering data corresponding to each cluster number within the range of cluster numbers, and take the initial clustering data corresponding to the cluster number with the best evaluation result as the clustering result.

[0031] S24. If the virtual object has a preset gain scheme at the first level, the clustering result is adjusted according to the preset gain scheme. The preset gain scheme includes adding a preset skill or increasing the value of attribute data at a preset rate.

[0032] S25. Extract the data features of each type from the clustering results, and establish an evaluation model that associates the first attribute data and the first interaction behavior data by combining the data features of each type.

[0033] Specifically, in processing pet data, in addition to routine cleaning operations, more detailed processing was performed based on the characteristics of the pet data. For example, when processing pet level-up data, it is checked whether the attribute growth values ​​conform to the normal growth range of that type of pet. If a pet's attack power increases far beyond the historical data range of pets of the same type at a certain level, it may be due to a temporary increase caused by a special event in the game (such as obtaining a powerful item). Such data will be specially marked or corrected when analyzing the growth curve to avoid misleading the clustering results. This adjustment allows subsequent clustering analysis to focus more on the inherent laws of pet's natural growth and combat performance, improving the accuracy of the clustering results in assessing the pet's true ability.

[0034] For pet data, the standardization process not only considers unit of measurement but also weights pet attributes based on their importance in battle. For example, in some pet battle games, attack power and defense power have different impacts on battle results; attack power may directly affect damage output, while defense power affects survivability. Therefore, during standardization, attack power is given a higher weight, while defense power is given a relatively lower weight, making the standardized data more reflective of the actual role of attributes in battle. This adjustment allows for better highlighting of attributes with a greater impact on battle results during cluster analysis, making the clustering results more consistent with the differences in pet performance in actual battles. This provides support for building a more accurate correlation model between attributes and battle performance.

[0035] In pet data clustering analysis, a method for dynamically determining the number of clusters, k, is employed. First, preliminary analysis of pet growth curve data and combat data is conducted to calculate data dispersion and similarity indices. For example, the Euclidean distance between attribute values ​​of different pets is calculated, and an initial range of k values ​​is determined based on the distance distribution. Then, within this range, various evaluation indicators, such as the silhouette coefficient and the Calinski-Harabasz index, are used to evaluate the clustering results under different k values. The k value that optimizes the evaluation indicators is selected as the final number of clusters. This method of dynamically adjusting the number of clusters, k, better adapts to the complexity and diversity of pet data, avoiding the problem of clustering results failing to accurately reflect pet data characteristics due to a fixed k value. For example, different types of pets have significantly different attribute distributions and growth patterns. Using a fixed k value might incorrectly cluster pets with different growth and combat performance patterns into one category, while dynamically determining the k value can more accurately classify pets, providing a more reliable foundation for subsequent analysis.

[0036] During the clustering process, expertise from the pet-raising domain is incorporated to optimize the clustering results. For example, if it's known that certain pets unlock special skills or experience accelerated attribute growth at specific level stages, the growth data and combat performance at these stages are considered as special factors in the clustering analysis. For pets with similar growth stage characteristics, even if their attribute values ​​are initially far apart in the clustering algorithm's calculations, the clustering results are adjusted appropriately based on domain knowledge to make them more likely to be grouped together. This allows for the discovery of pet data feature patterns based on the game's inherent mechanics, rather than solely relying on numerical similarity. This adjustment makes the clustering results more aligned with the actual situation of pet raising, providing players with more targeted training suggestions. For instance, players can use the clustering results to learn about the training experiences of other pets in the same category at specific stages, allowing them to better plan their own pet's training path. S3. In response to the interactive operation on the virtual object, generate a first attribute adjustment scheme for the virtual object according to the evaluation model.

[0037] If the interaction with the virtual object is a continuous attack operation that exceeds a preset number of times and a preset damage value, then the evaluation model is used to analyze the reference defense attribute data corresponding to the interaction operation. If the defense attribute value of the virtual object is less than the reference defense attribute data, then based on the difference between the defense attribute value of the virtual object and the reference defense attribute data, virtual items and virtual tasks are selected to improve the defense attribute value of the virtual object, and the virtual items and virtual tasks are used as the first attribute adjustment scheme of the virtual object.

[0038] Specifically, when a pet suffers continuous high damage in battle, the system responds quickly within 0.5 seconds and provides suggestions on increasing its defense attribute based on the evaluation model, including recommended defense training items and suitable defense training dungeons.

[0039] In this embodiment, a real-time monitoring thread is set up in the program (e.g., a thread created using Python's threading module), which performs a task to check the pet's combat status data every 0.1 seconds. When it is detected that the pet has suffered continuous high damage (determined by setting the number of consecutive damages and a damage threshold), the evaluation model is called through the model inference engine (e.g., using the predict method of the Scikit-learn model) to quickly calculate suggestions for training the defense attribute, specifically: (1) The evaluation model uses machine learning algorithms to perform in-depth analysis of the input data. The model compares the combat data and growth curves of a large number of pets of the same type to find the correlation patterns between attributes and combat performance. When analyzing the defense attribute, the model focuses on factors related to defense, such as the changes in the combat performance of other pets of the same type after increasing their defense when they receive a certain type of damage. For example, the analysis found that when a certain type of pet faces magic damage, increasing its magic defense attribute significantly reduces the damage received and increases its win rate. Through this correlation analysis, the model can determine the relationship between the current pet's defense attribute and combat performance, and judge whether its defense is insufficient.

[0040] (2) Based on the results of the correlation analysis, the evaluation model calculates the current pet's defensive attribute improvement needs. This may involve calculating the required defensive value and determining the type of defensive attribute that needs to be prioritized (physical defense, magical defense, or other special defenses). For example, if the model finds that the pet suffers significant damage when continuously hit by magical attacks, and that similar pets can effectively reduce damage after reaching a certain level of magical defense, then the model will calculate the required magical defense value for the current pet. This step clarifies the specific goals of pet defense development, making the development recommendations more targeted.

[0041] (3) Using content-based recommendation algorithms (such as implementing a content-based collaborative filtering recommendation algorithm using the Surprise library in Python), suitable defensive training items and defensive training dungeons are selected based on the pet's current attributes and combat situation. The model queries the item database for items that meet the requirements for increasing defense and matches dungeons with difficulty and training objectives from the dungeon database. For example, if the model calculates that the pet needs to increase magic defense, it will recommend items with attributes that increase magic defense, such as magic shield scrolls, and at the same time recommend training dungeons with the main goal of increasing magic defense, such as magic defense trial dungeons. These suggestions provide players with specific training paths, making it convenient for players to quickly take action to improve their pet's defense.

[0042] Specifically, features are extracted and encoded for information related to pet attributes, battle situations, items, and dungeons. For pet attributes, their numerical features are normalized to ensure they fall within a uniform range, facilitating subsequent calculations. For textual descriptions of items and dungeons, such as the applicable scenarios for items and the training objectives of dungeons, natural language processing techniques are used for keyword extraction and vectorization. For example, "items that increase magic defense" are represented as a numerical representation containing relevant feature vectors so that the algorithm can process and analyze them.

[0043] Based on the preprocessed data, an item feature matrix is ​​constructed: each row of the matrix represents an item, and each column represents a feature dimension, such as defense bonus, applicable pet type, and item rarity. This matrix clearly represents the feature information of each item, providing a foundation for subsequent calculations of the similarity between item and pet requirements. A dungeon feature matrix is ​​then constructed: rows represent different dungeons, and columns represent dungeon features, such as dungeon difficulty level, primary training defense attribute type, and dungeon rewards. In this way, the characteristics of each dungeon can be presented in matrix form, facilitating algorithm analysis and comparison.

[0044] Calculate the similarity between items and pet needs: For each pet, determine its need feature vector for defensive items based on its current attributes and combat situation. For example, if a pet takes a lot of magic damage in battle, its need for magic defense items will be higher, and the weight of the magic defense-related dimension in its need feature vector will increase accordingly. Then, use similarity calculation methods such as cosine similarity and Euclidean distance to calculate the similarity between the feature vector of each item and the pet's need feature vector. The higher the similarity, the more the item matches the pet's current defensive needs.

[0045] Calculate the similarity between the copy and the pet's needs: Similarly, determine the pet's required feature vector for defense training copies based on the pet's situation. For example, if the pet's current defense is low and it needs to comprehensively improve its defense capabilities, its required feature vector will focus more on copies of comprehensive defense training. Then, calculate the similarity between the feature vector of each copy and the pet's required feature vector to measure the degree of match between the copy and the pet's current defense training needs.

[0046] Filtering Defensive Training Items: Based on the calculated similarity between the items and the pet's needs, items are sorted from highest to lowest similarity. A recommendation threshold is set, and items with similarity higher than the threshold are selected as defensive training items to recommend to the player. For example, items in the top 20% with a similarity greater than 0.8 are recommended. These items are highly matched to the pet's current defensive needs, allowing players to more effectively improve their pet's defense.

[0047] Filtering Defense Training Dungeons: Similarly, dungeons are sorted based on their similarity to the pet's needs, with dungeons showing the highest similarity being recommended. The pet's current level and strength are also considered during dungeon selection to ensure the recommended dungeons are of moderate difficulty—neither too difficult for the pet to complete training nor too easy to achieve the desired training effect. For example, a lower-level pet will be recommended a less difficult dungeon that effectively improves its basic defense capabilities; a higher-level pet will be recommended a more challenging dungeon to further enhance its defense.

[0048] The recommendation process also considers player preferences and historical behavior data. If a player has historically preferred using certain types of items or participating in dungeons of a particular difficulty level, then those items and dungeons will receive a higher recommendation weight. For example, if a player frequently uses rare items to train their pet, then when recommending defensive training items, rarer items will be prioritized to satisfy the player's preferences. Since pet attributes and combat situations are constantly changing, the recommendation algorithm updates data in real time and dynamically adjusts the recommendations. If a pet is injured in battle, resulting in reduced health, or acquires a new skill that affects its combat performance, the algorithm recalculates the similarity and updates the recommended items and dungeons. This ensures that players always receive recommendations that best suit their pet's current situation, improving the efficiency and effectiveness of pet training.

[0049] (4) The calculated defense attribute training suggestions will be fed back to the player. During the player's implementation of the training plan, the game system will continuously collect data on the pet's attribute changes and combat performance. This new data will be input into the evaluation model again, and the model will optimize and adjust the training suggestions based on the new data. If the pet does not significantly reduce the damage it receives in subsequent battles after the player uses a certain defense training item as suggested, the evaluation model will reanalyze the data and adjust the training suggestions, such as changing the training item or adjusting the difficulty of the training dungeon, to ensure that the training plan can always effectively improve the pet's defense and adapt to the ever-changing combat environment.

[0050] S4. In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

[0051] In this embodiment, a data interaction interface is established with other pet-raising games, using either a RESTful API or the gRPC protocol. If a RESTful API is used, an API endpoint is created on the backend server using a framework (such as Node.js's Express framework or Python's Flask framework), defining the format of data requests and responses. If the gRPC protocol is used, the corresponding gRPC development tools (such as the @grpc / grpc-node library in Node.js) are used to define the service interface and message structure. A smart contract is deployed using a blockchain platform (such as the Ethereum consortium blockchain). First, a smart contract is written using the Solidity language, defining the encryption and signing rules for data transmission. For example, Ethereum's encryption functions are used to encrypt the data, and a signing function is used to sign the data, ensuring data security and immutability. The smart contract is deployed on the Ethereum consortium blockchain, providing the contract address and related interface information to other game developers. This enables the sharing of pet attribute data, training history data, and battle data between different games.

[0052] S41. In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

[0053] S42. If the first attribute growth curve and the second attribute growth curve have a similarity interval greater than a first threshold, and the similarity between the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval is greater than a second threshold, then a second attribute adjustment scheme for the virtual object is generated based on the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval.

[0054] Specifically, big data analytics is used to integrate data from multiple games. For example, analysis reveals that pets with a certain attribute have a higher win rate in PvP battles when their agility attribute reaches a certain value across multiple games. Based on this, players are provided with cross-game agility attribute training strategies, including general upgrade paths and skill combination suggestions. This allows players to develop scientific training strategies based on broader data, overcoming the limitations of single-game data and improving pet training success rates and combat strength.

[0055] This process involves reading data from a cross-game data warehouse and performing data cleaning, transformation, and loading (ETL) using big data processing frameworks such as Apache Hadoop and Spark. In Hadoop, the MapReduce programming model is used for initial data cleaning and transformation, removing duplicate and erroneous data. In Spark, the DataFrame and Dataset APIs are used for further data processing, converting the data into a format suitable for analysis.

[0056] Data mining algorithms (such as the Apriori association rule mining algorithm) were used to analyze data on pets of the same attribute in different games. In Python, the mlxtend library was used to implement the Apriori algorithm, setting appropriate support and confidence thresholds to identify general patterns in attribute training and skill combinations. For example, it was found that in multiple games, pets with a certain attribute, when their intelligence attribute reaches a certain value, can increase their magic damage output when paired with specific skills. These patterns were then transformed into general training strategies and stored in a strategy database (such as using MySQL to store strategy information) for players to query and use.

[0057] This solution involves cross-game pet training. While different games have their own characteristics, pets with the same attributes may share certain commonalities. By mining data on pets with the same attributes across different games, these common patterns can be discovered, allowing for the development of universal attribute training and skill combination strategies. For example, in multiple pet training games, there may be a pattern where, when a pet's agility attribute reaches a certain value, combining it with specific control skills can increase the win rate in PVP battles. This universal cross-game strategy cannot be implemented through the pre-set code of a single game. It provides players with a unified training approach when switching between different games, saving them the time and effort of figuring out training strategies anew. Furthermore, uncovering cross-game skill combination patterns helps promote communication and integration between different games. Game developers can learn from excellent skill combinations and attribute training methods in other games to enrich their own gameplay and strategies. Players can also apply training experience gained in one game to other games, enhancing player engagement and stickiness across different games. For example, a game can refer to popular skill combinations in other games to develop similar gameplay or introduce new pet skills, bringing more novelty and fun to players.

[0058] In some embodiments, data on players' spending habits (monthly recharge amount, type of items purchased) and game time (weekly online time) in different games are collected. This data is sent to the server through the communication interface between the game client and the server, and stored in a player profile database (such as using MongoDB to store player profile data).

[0059] Players are classified using clustering analysis algorithms (such as the DBSCAN algorithm, implemented in Python using the Scikit-learn library). First, player data is formatted suitable for the DBSCAN algorithm, with appropriate neighborhood radius and minimum number of points set. The DBSCAN algorithm is then run to classify players into different types, such as "casual players" and "heavy spenders."

[0060] For different types of players, suitable strategies are selected from a strategy database (which covers various strategies for pet attribute training, skill combinations, and resource acquisition, providing rich material for selecting suitable strategies). These strategies are then combined with the player's actual situation in each game (such as current pet level, attributes, and owned items) to generate a personalized cross-game training plan using a rule engine (e.g., the Drools rule engine in Java). This personalized training plan is then displayed to the player through the communication interface between the game client and server. For example, for "casual players," a training plan focusing on acquiring resources through daily tasks is recommended based on their playtime and spending habits, along with specific upgrade and skill learning steps tailored to the player's current pet's situation.

[0061] For example, after initially filtering strategies based on player type, the system further searches the strategy database from multiple dimensions. If a player has a pet with a specific attribute, the system searches for training strategies related to that attribute; if a player is at a specific level in a game, the system filters out strategies suitable for that level. For instance, if a player's pet is high-agility and they are at a mid-game level, the system will search the strategy database for training strategies suitable for high-agility pets in the mid-game, including skill combination suggestions and attribute enhancement priorities, ensuring that the selected strategies align with the player's current game situation and improving the strategies' practicality and effectiveness.

[0062] Different games differ in how they acquire and consume resources for pet training. Cross-game training allows players to integrate resources from multiple games, avoiding redundant investment in a single game. For example, specific items accumulated by a "casual player" in game A can be used for pet training in game B without needing to acquire them again in game B, saving time and effort and improving resource utilization efficiency. This is based on cluster analysis of player spending habits and playtime, allowing for customized training plans for different player types, rational resource allocation, and maximizing resource value.

[0063] In some embodiments, the pet development progress of the same player in the first application database and the second application database are interconnected. Pets developed to a certain stage in the first application can continue their development in the second application, maintaining a consistent development experience. Furthermore, the attribute adjustment directions differ between the first and second applications. Players can focus on pet attribute development in the first application and on skill combinations in the second application, enriching the player's gaming experience and increasing the game's fun and appeal.

[0064] In some embodiments, the system includes a player community integrating forums, live streaming, and friend interaction. Players can share pet training tips and combat experiences in the forum through text, pictures, and videos; showcase their pets' combat processes and training results through live streaming; and communicate with friends via private messages to share exclusive training techniques. The system uses natural language processing technology to perform semantic analysis on the text content in the forum, extracting key training information; and uses image recognition and video analysis technology to interpret information such as pet attributes and skill releases in live streams and shared videos. When a player shares their unique pet skill combination that achieved a high win rate in a specific dungeon, the system completes the analysis within one hour, incorporating information such as the applicable scenarios for the skill combination and the pet attribute requirements into the recommendation model. Based on the player's activity in the community, such as the number of posts, the quality of replies, and the number of viewers in the live stream, as well as the quality of the shared content, an AI-invited panel of professional game players and developers scores the player. After comprehensive evaluation, the player is provided with exclusive rewards such as rare pet eggs, limited-edition training items, and special titles to incentivize active participation in community interaction.

[0065] Specifically, a data collector is set up on the backend server, using a scheduled task tool (such as the node-schedule library in Node.js) to periodically retrieve data from the forum database (e.g., using MySQL to store forum data), the live video storage directory (e.g., the video path stored in MinIO), and the user chat history database (e.g., using MongoDB to store chat history). Natural language processing libraries (such as NLTK) are then used to process the forum text content. First, NLTK's tokenizer (e.g., the word_tokenize function) is used to segment the text, breaking sentences down into individual words. Then, a part-of-speech tagging tool (e.g., the pos_tag function) is used to tag each word with its part of speech, indicating whether it is a noun, verb, adjective, etc. Finally, a named entity recognition tool (e.g., the ne_chunk function) is used to identify key entities in the text (e.g., pet names, skill names, etc.) and extract key training information.

[0066] Image recognition libraries (such as OpenCV) and video analysis tools (such as FFmpeg combined with deep learning models) are used to extract and analyze information such as pet attributes and skill releases from live streams and shared videos. For example, OpenCV's object detection algorithms (such as cascaded classifiers based on Haar features) are used to detect pets in videos. By analyzing the pet's behavior and visual changes in the video, combined with deep learning models (such as action recognition models based on convolutional neural networks), skill release actions are identified, and pet attribute and skill information is extracted. The analysis results are stored in a knowledge graph database (such as Neo4j), and Neo4j's Cypher query language is used to create nodes and relationships, establishing knowledge associations (pet attributes are associated with combat performance, pet skills are associated with attributes and combat performance, pet training methods are associated with attribute improvement and combat performance, player behavior is associated with pet training, and pet training data is associated between different games), facilitating subsequent queries and recommendations.

[0067] A scoring engine is set up on the backend server to calculate a player's overall score based on their activity data in the community (number of posts, quality of replies, number of viewers in live streams, etc.) and the quality score of their shared content (scored by a panel of judges). For example, the weights for post quantity (0.3), reply quality (0.3), live stream viewership (0.2), and shared content quality score (0.2) are set. The overall score is calculated using the formula: "Overall Score = Number of Posts × 0.3 + Reply Quality × 0.3 + Live Stream Viewers × 0.2 + Shared Content Quality Score × 0.2".

[0068] Based on the score threshold, rewards such as exclusive rare pet eggs, limited-edition training items, and special titles are selected from the reward database (e.g., using MySQL to store reward information). Push notification code is written using the SDK of a push notification system (such as JPush) to send the reward information to the player. For example, in Node.js, using the official JPush SDK, the push interface is called to send reward notifications to the player's game client.

[0069] In a second embodiment of the present invention, in response to an operation of viewing attribute data for a virtual object, the following steps are further included: Time series analysis is used to analyze the first interaction behavior data and first attribute data of the virtual objects in the first application database within a preset time period to generate an attribute change prediction model. The attribute change prediction model is used to predict the change trend of the first attribute of the virtual object; By combining the corresponding environmental data and corresponding user data of the virtual object, the change trend of the first attribute is adjusted to obtain the change trend of the second attribute; Based on the changing trend of the second attribute, a periodic third attribute adjustment scheme is generated for the virtual object.

[0070] Specifically, time series analysis algorithms (such as the ARIMA model, implemented using the statsmodels library in Python) are used to process pet attribute changes and combat performance data. First, historical data is organized into a time series format. Then, the ARIMA model is used for fitting and prediction, forecasting growth trends for the next 10 or even 20 levels. The prediction results are divided into time periods (weekly) to generate long-term training plans of 30 days or more. These plans are then stored in the game server's planning database using database manipulation statements (such as SQL's INSERT INTO statement), and players can view them at any time through the game client's query interface.

[0071] In this embodiment, in addition to the pet's own data, in-game environmental data will also be collected, including the difficulty level and reward mechanism of different dungeons, the acquisition methods and attribute bonuses of various training items, as well as the popular combat strategies and mainstream pet training directions in the current game version. This environmental data reflects the overall ecosystem of the game and is crucial for formulating training plans that conform to the actual situation of the game.

[0072] Simultaneously, it collects player behavior data in the game, such as player playtime, game modes participated in, spending habits, and preferences for different pets. The inclusion of player data allows training plans to better align with players' actual gaming styles and levels of engagement, improving the feasibility and practicality of the plans.

[0073] The prediction results are adjusted by combining collected environmental and player data. If the game is about to update with a new dungeon that requires a high agility attribute for pets, the weight of agility improvement will be appropriately increased when predicting pet growth. If the player is a "casual player" with limited game time, the prediction model will be more inclined to plan ways to improve pet attributes through simple daily tasks, rather than complex and high-difficulty dungeon challenges.

[0074] Based on pet growth predictions, the game is divided into weekly cycles to determine weekly training priorities. For example, if a pet's defense is predicted to grow slowly over the next two weeks, but a new event dungeon requiring high defense is about to launch, then the focus for the first two weeks will be on increasing the pet's defense. This includes recommending the use of defense-boosting training items and participating in defense training dungeons. Specific expected attribute values ​​are set for each week's training priorities. These values ​​are determined based on the pet's growth predictions, game environment requirements, and player goals. For example, it is expected that by the end of the third week, the pet's physical defense should reach a certain value to meet the requirements for participating in a specific dungeon, while also aligning with the pet's overall growth plan. Considering player spending habits and in-game resource acquisition methods, the weekly pathways to obtain training resources are planned. For "casual players," it is recommended to acquire resources by completing daily tasks and participating in simple activities; for "heavy spenders," it is recommended to combine recharge discounts and limited-time events to plan the purchase of cost-effective training items. Additionally, based on the in-game dungeon reward mechanism, players will be recommended to participate in dungeons that provide the necessary training resources.

[0075] The established long-term training plan is validated using simulation algorithms. In the simulated environment, pets are trained according to the planned methods, and the changes in pet attributes are observed to ensure they meet expectations and perform ideally in different game scenarios (such as PVP battles and dungeon challenges). If the simulation results show that the pet performs poorly in certain scenarios or fails to reach the expected attribute values, the plan will be adjusted. Potential risks during the training planning process are assessed, such as game version updates causing changes to training items or dungeon mechanics, or players being unable to follow the plan due to real-world reasons. Corresponding countermeasures are developed for different risks. For example, if a game version update is predicted to affect the effect of a certain training item, alternative training schemes are planned in advance; for potential changes in player playtime, the plan includes some flexibility, allowing players to adjust their training progress within a certain range.

[0076] The generated long-term training plans are stored in the game server's planning database using database manipulation statements (such as SQL's INSERT INTO statement). The planning database categorizes and stores training plans for different players for easy management and retrieval. The stored data includes detailed information such as weekly training priorities, expected attribute values, and resource acquisition methods, ensuring data integrity and accuracy. A query interface is provided in the game client, allowing players to view their pets' long-term training plans at any time. The query interface is designed with user experience in mind, displaying training plans in an intuitive and easy-to-understand way, such as showing weekly training tasks and expected goals in a calendar format, helping players understand their training progress and direction, and rationally allocate their game time.

[0077] Please refer to Figure 2 Embodiment 3 of the present invention is as follows: A virtual object attribute adjustment scheme generation system 1 includes a memory 2, a processor 3, and a computer program stored on the memory 2 and executable on the processor 3. When the processor 3 executes the computer program, it implements the various steps of a virtual object attribute adjustment scheme generation method according to Embodiment 1 or 2.

[0078] In summary, the present invention provides a method and system for generating attribute adjustment schemes for virtual objects. By storing the first interaction behavior data and attribute growth values ​​of virtual objects in a first application database, it can dynamically track the growth process of virtual objects and establish attribute growth curves based on actual data, thereby providing data support for subsequent intelligent adjustments. By establishing an evaluation model and correlating attribute data with interaction behavior performance, it can accurately analyze the potential and shortcomings of virtual objects and provide users with targeted attribute adjustment schemes. By integrating virtual object data shared in a second application database, it can provide users with universal attribute adjustment schemes across applications, avoiding redundant training, saving cross-application training resources, and enhancing the linkage between applications and user experience. Furthermore, the introduction of time series analysis and environment / user data adaptation makes attribute prediction more accurate and periodic adjustment schemes more operable. By optimizing the model through clustering and pre-set gain schemes, the adaptability and recommendation accuracy of the evaluation model are improved. Defense enhancement schemes are quickly generated for high-damage scenarios, enhancing real-time strategy in combat. Generating universal schemes based on cross-application data similarity expands the applicability of training strategies and reduces user learning costs.

[0079] 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 generating attribute adjustment schemes for virtual objects, characterized in that, Including the following steps: The first interactive behavior data of the virtual object and the growth value of the first attribute data of the virtual object when the level data changes are stored in the first application database in real time. The growth curve of the first attribute data of the virtual object is established based on the growth value of the first attribute data. Based on the first interaction behavior data and the first attribute growth curve of the same type of virtual objects, an evaluation model is established that associates the first attribute data and the first interaction behavior data. In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model; In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

2. The method for generating an attribute adjustment scheme for a virtual object according to claim 1, characterized in that, In response to an operation to view attribute data of the virtual object, the method further includes: Time series analysis is used to analyze the first interaction behavior data and first attribute data of the virtual objects in the first application database within a preset time period to generate an attribute change prediction model. The attribute change prediction model is used to predict the change trend of the first attribute of the virtual object; By combining the corresponding environmental data and corresponding user data of the virtual object, the change trend of the first attribute is adjusted to obtain the change trend of the second attribute; Based on the changing trend of the second attribute, a periodic third attribute adjustment scheme is generated for the virtual object.

3. The method for generating an attribute adjustment scheme for a virtual object according to claim 1, characterized in that, Based on the first interaction behavior data and the first attribute growth curve of virtual objects of the same type, an evaluation model is established that associates the first attribute data and the first interaction behavior data, including: Calculate the first dispersion and first similarity of the first interaction behavior data of virtual objects of the same type, and calculate the second dispersion and second similarity of the first attribute growth curve of virtual objects of the same type. The range of cluster numbers is determined based on the first degree of dispersion, the first similarity, the second degree of dispersion, and the second similarity; The initial clustering data corresponding to each cluster number within the range of cluster numbers are evaluated, and the initial clustering data corresponding to the cluster number with the best evaluation result is taken as the clustering result; If the virtual object has a preset gain scheme at the first level, the clustering result is adjusted according to the preset gain scheme. The preset gain scheme includes adding a preset skill or increasing the value of the attribute data at a preset rate. Extract data features of each type from the clustering results, and establish an evaluation model that associates the first attribute data and the first interaction behavior data by combining the data features of each type.

4. The method for generating an attribute adjustment scheme for a virtual object according to claim 1, characterized in that, In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model, including: If the interaction with the virtual object is a continuous attack operation that exceeds a preset number of times and a preset damage value, then the evaluation model is used to analyze the reference defense attribute data corresponding to the interaction operation. If the defense attribute value of the virtual object is less than the reference defense attribute data, then based on the difference between the defense attribute value of the virtual object and the reference defense attribute data, virtual items and virtual tasks are selected to improve the defense attribute value of the virtual object, and the virtual items and virtual tasks are used as the first attribute adjustment scheme of the virtual object.

5. The method for generating an attribute adjustment scheme for a virtual object according to claim 1, characterized in that, Based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a general second attribute adjustment scheme for the virtual object is established, including: Based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a second attribute growth curve of the shared virtual object is generated. If the first attribute growth curve and the second attribute growth curve have a similarity interval greater than a first threshold, and the similarity between the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval is greater than a second threshold, then a general second attribute adjustment scheme for the virtual object is generated based on the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval.

6. A system for generating attribute adjustment schemes for virtual objects, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: The first interactive behavior data of the virtual object and the growth value of the first attribute data of the virtual object when the level data changes are stored in the first application database in real time. The growth curve of the first attribute data of the virtual object is established based on the growth value of the first attribute data. Based on the first interaction behavior data and the first attribute growth curve of the same type of virtual objects, an evaluation model is established that associates the first attribute data and the first interaction behavior data. In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model; In response to the attribute data viewing operation of the virtual object, a general second attribute adjustment scheme for the virtual object is established based on the second interaction behavior data, second attribute data and historical attribute adjustment scheme of the virtual object shared in the second application database.

7. The virtual object attribute adjustment scheme generation system according to claim 6, characterized in that, In response to an operation to view attribute data of the virtual object, the method further includes: Time series analysis is used to analyze the first interaction behavior data and first attribute data of the virtual objects in the first application database within a preset time period to generate an attribute change prediction model. The attribute change prediction model is used to predict the change trend of the first attribute of the virtual object; By combining the corresponding environmental data and corresponding user data of the virtual object, the change trend of the first attribute is adjusted to obtain the change trend of the second attribute; Based on the changing trend of the second attribute, a periodic third attribute adjustment scheme is generated for the virtual object.

8. The virtual object attribute adjustment scheme generation system according to claim 6, characterized in that, Based on the first interaction behavior data and the first attribute growth curve of virtual objects of the same type, an evaluation model is established that associates the first attribute data and the first interaction behavior data, including: Calculate the first dispersion and first similarity of the first interaction behavior data of virtual objects of the same type, and calculate the second dispersion and second similarity of the first attribute growth curve of virtual objects of the same type. The range of cluster numbers is determined based on the first degree of dispersion, the first similarity, the second degree of dispersion, and the second similarity; The initial clustering data corresponding to each cluster number within the range of cluster numbers are evaluated, and the initial clustering data corresponding to the cluster number with the best evaluation result is taken as the clustering result; If the virtual object has a preset gain scheme at the first level, the clustering result is adjusted according to the preset gain scheme. The preset gain scheme includes adding a preset skill or increasing the value of the attribute data at a preset rate. Extract data features of each type from the clustering results, and establish an evaluation model that associates the first attribute data and the first interaction behavior data by combining the data features of each type.

9. A virtual object attribute adjustment scheme generation system according to claim 6, characterized in that, In response to an interactive operation on the virtual object, a first attribute adjustment scheme for the virtual object is generated according to the evaluation model, including: If the interaction with the virtual object is a continuous attack operation that exceeds a preset number of times and a preset damage value, then the evaluation model is used to analyze the reference defense attribute data corresponding to the interaction operation. If the defense attribute value of the virtual object is less than the reference defense attribute data, then based on the difference between the defense attribute value of the virtual object and the reference defense attribute data, virtual items and virtual tasks are selected to improve the defense attribute value of the virtual object, and the virtual items and virtual tasks are used as the first attribute adjustment scheme of the virtual object.

10. A virtual object attribute adjustment scheme generation system according to claim 6, characterized in that, Based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a general second attribute adjustment scheme for the virtual object is established, including: Based on the second interaction behavior data, second attribute data, and historical attribute adjustment scheme of the virtual object shared in the second application database, a second attribute growth curve of the shared virtual object is generated. If the first attribute growth curve and the second attribute growth curve have a similarity interval greater than a first threshold, and the similarity between the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval is greater than a second threshold, then a general second attribute adjustment scheme for the virtual object is generated based on the first interaction behavior data and the second interaction behavior data corresponding to the similarity interval.

Images