Game account value retention prediction method and system based on artificial intelligence

Abstract

The invention provides a game account value retention prediction method and system based on artificial intelligence, and relates to the technical field of data processing, and the method comprises the steps: obtaining attribute data of a plurality of game accounts; preprocessing the attribute data; determining a clustering algorithm based on the preprocessed attribute data in combination with the type and the application scene of the game account; performing clustering analysis on the game account data through a clustering algorithm, and determining a standardized account category; acquiring historical price data of each game account in the standardized account category; constructing a time sequence data set according to the historical price data; inputting the time sequence data set into a pre-trained LSTM model, and outputting a price trend of the standardized account category; and according to the price trend and the real-time prices of the game accounts, determining the value retention of each game account. According to the invention, a more reliable reference standard is provided for a game account transaction market.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a game account value preservation prediction method and system based on artificial intelligence. BACKGROUND

[0002] Game account transactions, as an important derivative field of the network game industry, have rapidly grown with the booming development of popular games such as League of Legends, King's Glory, and Genshin Impact. Game accounts are no longer just a player's game identity identifier, but also carry a large amount of virtual asset value, including rare equipment, out-of-print props, and high-level characters. The value fluctuations of these virtual assets directly affect the market performance of game accounts, making game accounts have the attributes of financial products.

[0003] By predicting the value preservation of game accounts, a scientific, objective, and efficient value evaluation method can be provided for the rapidly growing game account transaction market, not only improving prediction accuracy, but also enhancing the comparability and reliability of the results. Its application can help players make rational transactions, reduce risks, provide intelligent pricing and risk control support for platforms, and promote the standardized management of virtual assets and the healthy development of the industry.

[0004] However, since game accounts are non-standardized products, each account's attribute combination has uniqueness, which poses a great challenge to value prediction. Traditional valuation methods are mainly based on current market prices and historical transaction records, but it is difficult to accurately predict the future price trend and value preservation potential of accounts, resulting in large deviations in prediction results, lack of credible reference for both parties in transactions, and thus possible market pricing chaos and increased user transaction risks. SUMMARY

[0005] In view of the above deficiencies of the prior art, the purpose of the embodiments of the present application is to provide a game account value preservation prediction method based on artificial intelligence, which can solve the technical problem that traditional valuation methods are mainly based on current market prices and historical transaction records, but it is difficult to accurately predict the future price trend and value preservation potential of accounts, resulting in large deviations in prediction results, lack of credible reference for both parties in transactions, and thus possible market pricing chaos and increased user transaction risks.

[0006] The first aspect of the embodiments of the present application proposes a game account value preservation prediction method based on artificial intelligence, comprising: S1: obtaining attribute data of a plurality of game accounts; S2: preprocessing the attribute data; S3: determining a clustering algorithm based on the preprocessed attribute data in combination with the type and application scenario of the game account; S4: performing clustering analysis on the preprocessed attribute data by the clustering algorithm to determine a standardized account category; S5: obtaining historical price data of each game account in the standardized account category; S6: constructing a time series data set according to the historical price data; S7: inputting the time series data set into a pre-trained LSTM model to output a price trend of the standardized account category; S8: determining a preservation property rating of each game account according to the price trend and a real-time price of the game account.

[0007] The second aspect of the embodiment of the application provides a game account preservation property prediction system based on artificial intelligence, which comprises a processor and a memory. The memory stores programs or instructions executable on the processor, and the programs or instructions are executed by the processor to implement the steps of the game account preservation property prediction method based on artificial intelligence according to the first aspect.

[0008] The third aspect of the embodiment of the application provides a readable storage medium, and the readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the game account preservation property prediction method based on artificial intelligence according to the first aspect.

[0009] The technical scheme provided by the embodiment of the application has at least the following beneficial effects: In the embodiment of the application, the non-standardized accounts are divided into standardized categories through preprocessing and clustering analysis of the game account attribute data, effectively solving the problem of large individual differences and difficulty in comparison of the accounts, and through obtaining and constructing the historical price time series data of the standardized categories, the pre-trained LSTM model is introduced for prediction, which can fully utilize the time series features to capture the price trend change rule, thereby improving the accuracy of future price trend prediction, and further combining the price trend and the real-time price to judge the preservation property of the account, not only providing an objective transaction reference for the user and reducing the risk caused by the valuation deviation, but also providing an intelligent pricing and risk control means for the transaction platform, and improving the standardization and credibility of the market. BRIEF DESCRIPTION OF DRAWINGS

[0010] The accompanying drawings are only for the purpose of illustrating specific embodiments and are not considered as limiting the application, and in the entire drawings, the same reference signs represent the same components. Obviously, the accompanying drawings in the following description are only some embodiments described in the embodiments of the application, and other drawings can be obtained according to the drawings for those skilled in the art without creative labor.

[0011] Figure 1is a flowchart of a game account value preservation prediction method based on artificial intelligence provided by an embodiment of the present application.

[0012] Figure 2 is a structural diagram of a game account value preservation prediction system based on artificial intelligence provided by an embodiment of the present application. DETAILED DESCRIPTION

[0013] In order for those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions of the present application will be described clearly and completely below in conjunction with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0014] The game account value preservation prediction method based on artificial intelligence provided by the embodiments of the present application will be described in detail below in conjunction with the drawings, specific embodiments and application scenarios.

[0015] Reference is made to the accompanying drawings Figure 1 , which shows a flowchart of a game account value preservation prediction method based on artificial intelligence provided by an embodiment of the present application.

[0016] The game account value preservation prediction method based on artificial intelligence provided by the embodiments of the present application can include the following steps: S1: Obtain attribute data of a plurality of game accounts.

[0017] Among them, the game account refers to a virtual identity registered and used in a specific network game, containing information such as the level, equipment, skills, rare items, and game duration of the player. The attribute data refers to various feature information that can represent the value of the game account, such as account level, equipment configuration, rare item possession, historical activity, and recharge record.

[0018] It should be noted that by systematically obtaining attribute data of a plurality of game accounts, it can be ensured that the data source is extensive and the dimension is comprehensive, thereby providing a solid foundation for subsequent preprocessing, clustering, and prediction models. Unlike relying on a single account sample or a small amount of data, batch data collection not only improves the generalization ability of the model, but also effectively reduces the bias caused by individual account characteristics.

[0019] In one possible implementation, the attribute data includes game account level information, game account equipment configuration, game account rare item possession, and game duration.

[0020] S2: Preprocess the attribute data.

[0021] Specifically, the preprocessing specifically includes: cleaning processing, standardization processing, and feature selection processing.

[0022] It should be noted that the cleaning, standardization and feature selection of the original data ensure the quality of subsequent clustering analysis.

[0023] S3: Based on the pre-processed attribute data, determine the clustering algorithm combined with the type of game account and application scenario.

[0024] The type of game account refers to different classification dimensions of the game account, such as distinguishing by game category (MMORPG, MOBA, shooting, etc.), by account characteristics (high-level number, equipment number, rare item number, ordinary number), etc. It is used to guide the direction of clustering analysis. The application scenario refers to the actual use environment faced by the account value prediction, such as second-hand transaction platform pricing, player buying and selling reference, platform risk control, virtual asset evaluation, etc. Different scenarios have different requirements for the fineness and adaptability of clustering results.

[0025] The clustering algorithm is an unsupervised learning method, which is used to group accounts according to the similarity of data, such as K-means clustering, DBSCAN clustering, etc. The similarity of accounts in the same group is high, and the difference between different groups is significant.

[0026] It should be noted that by combining the type of game account and the application scenario to select the appropriate clustering algorithm, the clustering process can be more targeted and adaptive, thereby effectively improving the rationality and practicality of the clustering results. Traditional methods often use a single algorithm, which is easy to fail when the account distribution is sparse or the noise is high. In this step, K-means or DBSCAN algorithms are selected flexibly according to the actual scene, which not only ensures the clustering accuracy of high-density accounts, but also handles different density or abnormal point accounts.

[0027] In one possible implementation, the clustering algorithm specifically includes: a DBSCAN clustering algorithm and a K-means clustering algorithm.

[0028] The processing mode of the DBSCAN clustering algorithm includes hierarchical processing mode and non-hierarchical processing mode.

[0029] Among them, DBSCAN clustering algorithm, full name Density-Based Spatial Clustering of Applications with Noise, that is, density-based clustering algorithm. It identifies clusters through density threshold and neighborhood radius, and can effectively discover clusters of arbitrary shape and automatically identify noise points. K-means clustering algorithm is a partition-based clustering algorithm, which divides data into K clusters through iterative optimization, minimizes the difference within the cluster, and maximizes the difference between the clusters. The advantage is high computational efficiency, suitable for large-scale data sets.

[0030] Among them, hierarchical processing mode refers to adjusting parameters (such as neighborhood radius ε and minimum point number MinPts) in DBSCAN algorithm layer by layer, realizing multi-level clustering of data points with different density levels, and better depicting complex data distribution. Non-hierarchical processing mode refers to that DBSCAN algorithm only clusters under a single parameter configuration, directly outputs the cluster division result, and is suitable for data sets with relatively uniform distribution.

[0031] Specifically, the hierarchical processing mode is applicable in the following cases: the data distribution has obvious hierarchical structure characteristics, the player behavior in the game has obvious level or stage difference, the player groups with different density levels need to be identified, or the data set size is large and the processing efficiency needs to be improved.

[0032] The non-hierarchical processing mode is applicable in the following cases: the data distribution is relatively uniform and has no obvious hierarchical structure, the game type is simple and the player behavior pattern is relatively uniform, the data set size is moderate, the computing resources are sufficient, or the requirement for the fineness of the clustering result is not high.

[0033] It should be noted that by simultaneously using K-means and DBSCAN clustering algorithms and setting hierarchical and non-hierarchical processing modes for DBSCAN, the clustering process has stronger flexibility and applicability. K-means algorithm can efficiently process large-scale, relatively uniform distributed data, while DBSCAN is good at discovering clusters of arbitrary shape and removing noise points. The hierarchical processing mode further enhances the adaptability of DBSCAN to multi-density data, while the non-hierarchical mode is suitable for quickly dividing data.

[0034] S4: performing clustering analysis on the preprocessed attribute data through the clustering algorithm to determine the standardized account category.

[0035] Among them, the standardized account category refers to the unified category formed by the clustering result, which divides the originally non-standardized and personalized accounts with obvious differences into several standard groups, facilitating subsequent comparison and prediction.

[0036] It's worth noting that cluster analysis categorizes non-standardized game accounts into standardized categories, effectively addressing the problem of large differences in account attribute combinations and the difficulty in direct comparison. In this way, the system can group accounts with similar characteristics into the same category, thereby capturing market patterns at the category level and reducing the noise impact of individual differences. This not only improves the accuracy of subsequent time series modeling and price trend prediction but also enhances the stability and reliability of account valuation.

[0037] In one possible implementation, the clustering process of the DBSCAN clustering algorithm specifically includes: Based on the preprocessed attribute data, the parameters of the DBSCAN clustering algorithm are initialized: in, This represents the neighborhood radius, which is a parameter of the DBSCAN clustering algorithm. Indicates the regulating factor. This represents the standard deviation of the preprocessed attribute data.

[0038] Based on the initialized parameters, calculate the local density of each game account: in, Indicates the first i Local density of individual game accounts Indicates the first i game account of Neighborhood set It represents pi (π).

[0039] Local density characterizes the density of the account in the data distribution.

[0040] Based on local density, cluster analysis is performed on the attribute data of game accounts to determine standardized account categories: in, Indicates the first h The neighborhood radius of the layer The decay factor representing the neighborhood radius. Indicates the first h The minimum neighborhood threshold for a layer. This represents the initial minimum number of points threshold. Indicates the first h The amplification factor of the layer.

[0041] It should be noted that the algorithm can identify different granularity clustering structures layer by layer, not only improving the rationality and stability of the clustering results, but also providing solid data support for the formation of standardized account categories. This improvement makes the algorithm more robust and efficient when facing complex and diverse account attribute data.

[0042] S5: Obtain historical price data of each game account in the standardized account category.

[0043] Among them, the historical price data refers to the transaction price, listing price or valuation record of the game account in the past period of time, reflecting the value change trajectory of the account in the market.

[0044] It should be noted that based on the standardized account category, the system collects the historical price data of all accounts in this category, and constructs a sufficient time series data set. Using these data to train the time series prediction model, the accurate prediction of future price trend is realized.

[0045] S6: Construct a time series data set according to the historical price data.

[0046] Among them, the time series data set refers to arranging the historical price data in chronological order, and constructing fixed length sliding window samples when necessary, to form the input data format that can be processed by the prediction model.

[0047] It should be noted that by constructing the historical price data into a time series data set, the dynamic characteristics of the price change over time can be fully retained, so that the model can capture the trend, periodicity and volatility rules. This structured processing converts scattered historical prices into learnable sequence input, avoiding the problem that single point data cannot reflect the overall trend.

[0048] In one possible implementation, after S6, it further includes: Normalizing the time series data set: Among them, denotes the normalized time series data set, denotes the time series data set, denotes the mean of the time series data set, denotes the standard deviation of the time series data set.

[0049] It should be noted that by normalizing the time series data set, the influence of the differences in dimension and value range between different account price data can be eliminated, so that the data of various types can be compared and modeled on the same scale, thereby avoiding the problem that high value features dominate the model training.

[0050] S7: inputting the time series dataset into a pre-trained LSTM model to output a standardized account category price trend.

[0051] wherein the pre-trained LSTM model refers to LSTM (Long Short-Term Memory), which is an improved recurrent neural network good at processing and predicting long-term dependencies in time series data. "Pre-training" means that the model has optimized parameters through a large amount of training data and can be directly used for prediction tasks. The price trend is the predicted direction and amplitude of the account price in the future period based on historical data, which may be represented as rising, falling or fluctuating.

[0052] It should be noted that by inputting the time series dataset into the pre-trained LSTM model, its ability to capture long-term dependencies and nonlinear features can be fully utilized, thereby more accurately predicting the future price trend of the account category. Compared with traditional statistical methods or shallow models, LSTM has stronger robustness and generalization ability in processing complex time series data, and can effectively identify trends, periodicity and abnormal fluctuations.

[0053] In one possible implementation, the LSTM model includes an input layer, a plurality of LSTM layers with residual connections, an attention mechanism layer, and a fully connected output layer connected in sequence.

[0054] wherein the LSTM layer is a long short-term memory network layer that can capture long-term dependencies in time series through a gating mechanism to extract dynamic features of price changes over time. The attention mechanism layer refers to assigning weights to hidden states at different time steps to highlight the contribution of key moments to prediction, thereby enhancing the model's attention to important features. The fully connected output layer refers to mapping the context information output by the attention layer and the hidden state to the final prediction result space to output the price trend.

[0055] It should be noted that by sequentially setting the input layer, LSTM layer with residual connection, attention mechanism layer and fully connected output layer in the LSTM model, the advantages of deep feature extraction and key information selection can be fully combined in structure. Residual connection solves the problem of gradient disappearance in deep network training, allowing the model to stack more layers to enhance feature expression capability, and attention mechanism allows the model to automatically focus on the most important time points in historical prices for future trends, improving the accuracy and interpretability of the prediction results.

[0056] In one possible implementation, S7 specifically includes.

[0057] S701: inputting the time series dataset into the input layer.

[0058] S702: Extracting the hidden states of time series data in the LSTM layer: in, Indicates time t Input time, Indicates time t The Gate of Oblivion This represents the sigmoid activation function. express The hidden state at all times Indicates time t Input gate at time, Indicates time t Input time series data, Indicates time t The possible new state at that time Represents the hyperbolic tangent function. Indicates time t Output gate at time, Indicates time t The cell states retained after considering past states and current candidate states. express The time series data input at time 10:00. This indicates that the LSTM network is in t The output at time t, i.e., the hidden state at time t. This represents the dot product operation. and Each represents a weight matrix used to control the inputs of each gate and the influence of past states on the current state. and Both represent bias terms used to adjust the gating and candidate state activation levels.

[0059] S703: By using residual connections, the hidden states of each LSTM layer are fused to obtain the fused state: in, Indicates the first l Layer in t The hidden state at any given moment, i.e., the merged state. Indicates the first l The LSTM network function of the layer, Indicates the first Layer in t The hidden state at any given moment.

[0060] S704: Determine the attention weights of the fusion state at each time step through the attention mechanism layer: in, expresst In the time series data at time point 1 i The relevance score of each hidden state This represents a learnable weight vector. Indicates transpose. Represents a non-linear activation function. and Each represents a learnable weight matrix. Indicating time series data i The hidden state at all times express t The hidden state at all times express t At that moment, the i Attention weights for each hidden state Represents an exponential function. express t In the time series data at time point 1 j The relevance score of each hidden state express t The context vector at time step, Indicates the first i A hidden state.

[0061] S705: Through a fully connected output layer, based on the fusion state and attention weights at each time step, it outputs the price trend of standardized account categories. in, express t The predicted output value at any given time, i.e., the price trend. This represents the weight matrix of the fully connected output layer. This represents the bias term of the fully connected output layer.

[0062] It's important to note that, firstly, the gating mechanism of the LSTM layer effectively extracts long-term dependencies in time series data, avoiding the gradient vanishing problem common in traditional RNNs. Secondly, the introduction of residual connections improves the training stability of deep LSTM networks, giving the model stronger feature extraction capabilities. Furthermore, the attention mechanism allows the model to automatically identify and focus on the most critical time segments for the prediction results, thereby improving prediction accuracy and interpretability. Finally, fully connected layers map the extracted features to the price trend output, making the results more intuitive and easier to apply. Overall, this structure enhances the model's expressive power and improves prediction accuracy and stability, providing strong support for game account price trend analysis and value preservation assessment.

[0063] In one possible implementation, the process of determining the pre-trained LSTM model specifically includes: Determine the loss function based on the price trend and the real-time price of the game account: wherein, represents the loss function, represents the mean square error, represents the adjustment coefficient, represents the regularization term.

[0064] wherein, the loss function (Loss) refers to a function that measures the difference between the model prediction value and the true value.

[0065] According to the loss function, dynamically adjust the learning rate: wherein, represents the learning rate at the i-th iteration, t represents the initial learning rate, represents the decay factor at the i-th iteration. t According to the dynamically adjusted learning rate, update the parameters of the LSTM model through the backpropagation algorithm until the value of the loss function is less than the preset loss function value, and determine the pre-trained LSTM model:

[0066] wherein, represents the model parameters at the i-th iteration, represents the model parameters at the i-th iteration, represents the gradient of the loss function under the current parameters . t wherein, the backpropagation algorithm refers to the core algorithm of deep learning training, which is used to calculate the gradient of the loss function with respect to the parameters and update the parameters. It should be noted that the training process of the LSTM model is specified as four links of loss function setting, dynamic learning rate adjustment, backpropagation optimization and convergence judgment, forming a complete parameter optimization closed loop. By introducing the regularization term in the loss function, the risk of overfitting is effectively reduced, and the model has better generalization ability. Through the dynamic learning rate decay mechanism, it ensures that the training converges quickly in the early stage and approaches the optimal solution stably in the later stage. Through gradient backpropagation and iterative update, the model can gradually optimize the parameters until the preset convergence condition is reached.

[0067]

[0068] It should be noted that the training process of the LSTM model is specified as four links of loss function setting, dynamic learning rate adjustment, backpropagation optimization and convergence judgment, forming a complete parameter optimization closed loop. By introducing the regularization term in the loss function, the risk of overfitting is effectively reduced, and the model has better generalization ability. Through the dynamic learning rate decay mechanism, it ensures that the training converges quickly in the early stage and approaches the optimal solution stably in the later stage. Through gradient backpropagation and iterative update, the model can gradually optimize the parameters until the preset convergence condition is reached.

[0069] S8: Determine the preservation rating of each game account based on the price trend and the real-time price of the game account.

[0070] ​​​The real-time price of the game account refers to the actual listing price or transaction price of the account on the market at the current time, reflecting the immediate market value. The preservation rating is used to measure the ability of the game account to maintain or improve its value in the future. If the predicted trend shows that the future price is higher or stable than the current price, the account has preservation value, otherwise there is a risk of devaluation.

[0071] Specifically, in combination with the game life cycle, version update and other factors, the preservation rating (preservation, non-preservation) is determined.

[0072] It should be noted that by combining the predicted price trend with the real-time price of the account, the dynamic and forward-looking evaluation of the individual account value can be achieved. Unlike relying solely on historical prices or static attributes, this method combines model prediction results with market immediate data, reflecting both long-term trends and current market conditions, making the judgment result more scientific and accurate.

[0073] The technical solutions provided by the embodiments of the present application have at least the following beneficial effects: In the embodiments of the present application, by preprocessing and clustering analysis of the game account attribute data, the non-standardized accounts are divided into standardized categories, effectively solving the problem of large individual differences and difficulty in comparison. By obtaining and constructing the historical price time series data of the standardized categories, the pre-trained LSTM model is introduced for prediction, which can fully utilize the time series features to capture the price trend change rule, thereby improving the accuracy of future price trend prediction. Further combining the price trend and real-time price to judge the preservation of the account not only provides an objective trading reference for users, reduces the risk of valuation deviation, but also provides intelligent pricing and risk control means for the trading platform, improving the standardization and credibility of the market.

[0074] Referring to the accompanying drawings Figure 2 , a structure schematic diagram of a game account preservation prediction system based on artificial intelligence provided by an embodiment of the present application is shown.

[0075] An embodiment of the present application provides a game account preservation prediction system based on artificial intelligence 20, comprising a processor 201 and a memory 202. The memory 202 stores programs or instructions that can run on the processor 201. The programs or instructions are executed by the processor 201 to implement the steps of the above-mentioned game account preservation prediction method based on artificial intelligence, and can achieve the same technical effect. To avoid repetition, the present application will not be described again.

[0076] It is to be understood that the processor 201 in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0077] It is also to be understood that the memory 202 in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memory. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example, but not limitation, many forms of random access memory are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRAM).

[0078] The above-described embodiments can be implemented in whole or in part by software, hardware (e.g., circuitry), firmware, or any combination of the three. When implemented in software, the above-described embodiments can be implemented in the form of one or more computer programs that are stored in a computer-readable storage medium. The computer-readable storage medium stores one or more computer instructions or computer programs that, when loaded into a computer, cause the computer to perform the processes or functions described in the embodiments of the present application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, such as from a website, a computer, a server, or a data center to another website, computer, server, or data center, via a wired (e.g., infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more collections of available media. The available media can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0079] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0080] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0081] Those of ordinary skill in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the devices, apparatuses and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0082] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other manners. For example, the embodiments of the apparatus described above are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0083] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0084] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit.

[0085] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts of the prior art that make contributions or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0086] The embodiments of the present application provide a readable storage medium, which includes a program or instructions stored thereon. The program or instructions are executed by a processor to implement the steps of the game account value preservation prediction method based on artificial intelligence described above, and achieve the same technical effects. To avoid repetition, the present application will not be described again.

[0087] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present application, but not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. Any changes or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application.

Claims

1. An artificial intelligence-based game account value preservation prediction method, characterized in that, The method comprises the following steps: S1: obtaining attribute data of a plurality of game accounts; S2: preprocessing the attribute data; S3: determining a clustering algorithm based on the preprocessed attribute data in combination with the type and application scenario of the game account; S4: performing clustering analysis on the preprocessed attribute data by using the clustering algorithm to determine a standardized account category; S5: obtaining historical price data of each game account in the standardized account category; S6: constructing a time series data set according to the historical price data; S7: inputting the time series data set into a pre-trained LSTM model to output a price trend of the standardized account category; S8: determining the value preservation rating of each game account according to the price trend and the real-time price of the game account. 2.The AI-based game account value preservation prediction method of claim 1, wherein, The attribute data includes game account level information, game account equipment configuration, game account rare item possession situation, and game duration. 3.The AI-based game account value preservation prediction method of claim 1, wherein, The clustering algorithm specifically includes a DBSCAN clustering algorithm and a K-means clustering algorithm. The processing mode of the DBSCAN clustering algorithm includes a hierarchical processing mode and a non-hierarchical processing mode. 4.The AI-based game account value preservation prediction method of claim 3, wherein, The clustering process of the DBSCAN clustering algorithm specifically includes: initializing parameters of the DBSCAN clustering algorithm based on the preprocessed attribute data; wherein, denotes the neighborhood radius, i.e. a parameter of the DBSCAN clustering algorithm, denotes an adjustment factor, denotes the standard deviation of the pre-processed attribute data; calculating the local density of each game account according to the initialized parameters; wherein, denotes the i local density of the game account, i denotes the neighborhood set of the game account, denotes the mathematical constant pi; performing clustering analysis on the attribute data of the game account according to the local density to determine the standardized account category; wherein, represents the h neighborhood radius of the layer, represents a decay factor for the neighborhood radius, represents the h minimum neighborhood point threshold of the layer, represents an initial minimum point threshold, amplification factor of the h layer. 5.The AI-based game account value preservation prediction method of claim 1, wherein, After S6, it further comprises: normalizing the time series data set; wherein, denotes the normalized time series dataset, denotes the time series dataset, denotes the mean of the time series dataset, denotes the standard deviation of the time series dataset. 6.The AI-based game account value preservation prediction method of claim 1, wherein, The LSTM model comprises an input layer, a plurality of LSTM layers with residual connection, an attention mechanism layer, and a fully connected output layer connected in sequence. 7.The AI-based game account value preservation prediction method of claim 6, wherein, S7 specifically comprises: S701: inputting the time series data set into the input layer; S702: extracting the hidden state of the time series data in the LSTM layer; wherein, denotes the input at time t denotes the forget gate at time t denotes the sigmoid activation function, denotes the hidden state at time denotes the input gate at time t denotes the time series data input at time t denotes the possible new state at time t denotes the hyperbolic tangent function, denotes the output gate at time t denotes the cell state preserved after considering past states and current candidate state at time t denotes the time series data input at time denotes the output of the LSTM network at time denotes the hidden state at time t, t denotes the dot product operation, and both denote weight matrices for controlling the influence of input and past state on current state for each gate, and both denote bias terms for adjusting the gate and candidate state activation levels;​​​​​​​ S703: fusing the hidden states of each LSTM layer through residual connection to obtain a fusion state: wherein, denotes the hidden state of the l layer at time step t, t denotes the hidden state, i.e. the fusion state, of the denotes the hidden state of the l layer at time step t, denotes the hidden state of the layer at time step t, t denotes the hidden state of the S704: determining the attention weight of the fusion state at each time point through the attention mechanism layer: where, represents t the relevance score of the i th hidden state in the time series data at time represents a learnable weight vector, represents transpose, represents a nonlinear activation function, and both represent a learnable weight matrix, represents the hidden state at time i in the time series data, represents t the hidden state at time t , the attention weight of the i th hidden state at time represents an exponential function, represents t the relevance score of the j th hidden state in the time series data at time represents t the context vector at time represents the i th hidden state; S705: outputting the price trend of the standardized account category according to the fusion state at each time point and its attention weight through the fully connected output layer: wherein, denotes t a predicted output value at a time instant, i.e. a price trend, denotes a weight matrix of the fully connected output layer, denotes a bias term of the fully connected output layer. 8.The AI-based game account value preservation prediction method of claim 1, wherein, The determination process of the pre-trained LSTM model specifically comprises: determining a loss function based on the price trend and the real-time price of the game account; wherein, represents a loss function, represents a mean squared error, represents an adjustment coefficient, represents a regularization term; dynamically adjusting the learning rate according to the loss function; in, Indicates the first t Learning rate at the next iteration This represents the initial learning rate. Indicates the first t The decay factor at the next iteration; updating the parameters of the LSTM model through the back propagation algorithm according to the dynamically adjusted learning rate until the value of the loss function is less than a preset loss function value, and determining the pre-trained LSTM model: where, denotes the model parameters at the th iteration, denotes the model parameters at the t th iteration, denotes the gradient of the loss function at the current parameters . 9.A system for predicting the value preservation of a game account based on artificial intelligence, characterized in that, It comprises: a processor and a memory; The memory stores programs or instructions that can be run on the processor, and the programs or instructions are executed by the processor to implement the steps of the game account value preservation prediction method based on artificial intelligence according to any one of claims 1 to 8.

10. A readable storage medium, characterized by, The readable storage medium stores programs or instructions, and the programs or instructions are executed by the processor to implement the steps of the game account value preservation prediction method based on artificial intelligence according to any one of claims 1 to 8.

Images