A game data mining method

Abstract

The present invention relates to the field of virtual reality technology, and in particular to a game data mining method, comprising step S1, collecting virtual reality game performance data; step S2, preprocessing the virtual reality game performance data to obtain preprocessed data; step S3, analyzing the performance bottleneck factors of the virtual reality game; step S4, inputting the performance bottleneck factors of the virtual reality game into an optimization decision model, and outputting a virtual reality game performance optimization strategy; step S5, judging the use of user behavior data according to the number of repetitions of user behavior data, and correcting the virtual reality game performance optimization process according to the judgment result; step S6, judging the correlation between the use of user behavior data and the performance bottleneck factors of the virtual reality game according to an association rule mining algorithm, and adjusting the correction virtual reality game performance optimization process according to the judgment result. The present invention improves the optimization efficiency of virtual reality game performance.

Description

Technical Field

[0001] The present invention relates to the field of virtual reality technology, and in particular to a game data mining method. Background Art

[0002] Virtual reality games involve complex graphics rendering, physical calculations and other aspects, and place higher demands on technologies such as game engines and rendering pipelines. Providing a smooth gaming experience with lower hardware configurations is crucial to maintaining player immersion. In virtual reality games, data mining technology can be used to gain an in-depth understanding of players' gaming habits, preferences and needs, thereby optimizing game content and improving gaming experience. Based on the results of data mining, virtual reality games can be optimized in a targeted manner.

[0003] Chinese Patent Publication No.: CN108170592A discloses a remote testing method and terminal for virtual reality software performance. The present invention mounts a preset test script to the virtual reality software; when receiving a test request sent by the server, the test script obtains all UI objects corresponding to the virtual reality software to obtain a first UI object set; the test script obtains the M i pMap function state corresponding to each UI object in the first UI object set to obtain a function state set; the M i pMap function state includes an on state and a off state; the test script sends the first UI object set and the function state set to the server, so that the server generates a test report based on the first UI object set and the function state set. However, the invention only provides virtual reality software performance testing and cannot improve the optimization efficiency of virtual reality game performance. Summary of the invention

[0004] To this end, the present invention provides a game data mining method to overcome the problem of low efficiency in virtual reality game performance optimization in the prior art.

[0005] To achieve the above object, the present invention provides a game data mining method, comprising:

[0006] Step S1, collecting virtual reality game performance data;

[0007] Step S2, preprocessing the virtual reality game performance data to obtain preprocessed data;

[0008] Step S3, analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data;

[0009] Step S4, inputting the virtual reality game performance bottleneck factor into the optimization decision model, and outputting the virtual reality game performance optimization strategy;

[0010] Step S5, judging the usage of the user behavior data according to the number of repetitions of the user behavior data, and correcting the virtual reality game performance optimization process according to the judgment result;

[0011] Step S6, judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, and adjusting the virtual reality game performance optimization process according to the judgment result.

[0012] Further, in step S2, the preprocessing method includes:

[0013] Step S20, cleaning the virtual reality game performance data to obtain valid data;

[0014] Step S21, dividing the valid data into image data and data point data;

[0015] Step S22, standardizing the data point data by using a standardization algorithm to obtain standardized data point data;

[0016] Step S23, extracting features from the image data and the standardized data point data, and merging the feature-extracted data to obtain preprocessed data.

[0017] Furthermore, in step S20, when the virtual reality game performance data is cleaned, the missing values ​​of the virtual reality game performance data are filled by the interpolation method, the abnormal values ​​of the virtual reality game performance data are detected by the IQR method, and the representative information of the abnormal values ​​is collected, and the importance of the representative information of the abnormal values ​​is judged. The abnormal values ​​are deleted and marked according to the judgment result to obtain valid data, wherein:

[0018] When the information represented by the outliers is important, keep and mark the outliers;

[0019] When the information represented by the outlier is unimportant, the outlier is deleted.

[0020] Further, in the step S21, the valid data stored in the form of pixels is divided into image data, and the valid data stored in the form of numerical values ​​is divided into data point data according to the storage form of the valid data;

[0021] In the step S22, the data point data is standardized by the Z-score standardization method in the standardization algorithm, and the calculation formula of the standardization algorithm is set to Z=(X-μ) / σ, where: X is the data point data, μ is the mean of the data point data, σ is the standard deviation of the data point data, and Z is the standardized data point data;

[0022] The calculation formula for setting the mean μ is Where N is the number of data points in the data point data, Xi is the i-th data point in the data point data, for example, the first data point in the data point data is X1, the second data point is X2, ... the i-th data point is Xi;

[0023] The calculation formula for setting the standard deviation σ is

[0024] Furthermore, in step S23, feature extraction is performed on the image data by using a Canny edge detection method to obtain feature image data, and the Canny edge detection method includes:

[0025] Step G100, using Gaussian filtering to remove noise from the image data to obtain image data after the noise is removed;

[0026] Step G200, using the Sobel operator to calculate the gradient direction of the image data after the noise is removed, to obtain the gradient direction of the image data;

[0027] Step G300, using a non-maximum suppression method to perform non-maximum suppression in the gradient direction of the image data, retaining local maximum points, and obtaining a maximum point distribution of the image data;

[0028] Step G400, using a double threshold processing method to judge each point in the maximum point distribution of the image data, and performing edge connection based on the judgment result, setting the value of each maximum point to V, the maximum threshold to Vmax, and the minimum threshold to Vmin, where:

[0029] When V>Vmax, the maximum point is determined to be a strong edge;

[0030] When Vmin≤V≤Vmax, the maximum point is determined to be a weak edge;

[0031] When V<Vmin, the maximum point is determined to be a non-edge point;

[0032] Step G500, connecting the strong edge and the weak edge, and outputting the connected image as feature image data.

[0033] Furthermore, in step S23, the standardized data point data is feature extracted by principal component analysis to obtain feature data point data, and the feature image data and feature data point data are input into a deep feature fusion model for merging, and the deep feature fusion model outputs preprocessed data.

[0034] Furthermore, in step S3, when analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data, the preprocessed data is input into the bottleneck factor analysis model, and the performance bottleneck factors of the virtual reality game are output through the bottleneck factor analysis model.

[0035] Furthermore, in step S4, an optimization decision model is constructed through the expert optimization strategy data set, and the virtual reality game performance bottleneck factor is input into the optimization decision model to output the virtual reality game performance optimization strategy.

[0036] Furthermore, in step S5, the user behavior data is obtained, the number of repetitions J of the user behavior data is compared with the preset number of repetitions J0 of the user behavior data, the use of the user behavior data is judged according to the comparison result, and the compliance of the virtual reality game performance optimization process is judged and corrected according to the judgment result, wherein:

[0037] When J<J0, the user behavior data is determined to be low-frequency user behavior data, the virtual reality game performance optimization process meets the standard, and the virtual reality game performance optimization process is not corrected:

[0038] When J≥J0, the user behavior data is determined to be high-frequency user behavior data, and the virtual reality game performance optimization process is not up to standard. The virtual reality game performance optimization process is corrected, and an expert optimization strategy data set update window is pushed to the virtual reality game device administrator. The virtual reality game device administrator updates the expert optimization strategy data set, obtains the updated expert optimization strategy data set, and optimizes the virtual reality game performance according to the updated expert optimization strategy data set.

[0039] Furthermore, in step S6, when judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, the confidence P of the user behavior data usage and the virtual reality game performance bottleneck factor is calculated according to the association rule mining algorithm, and the confidence P is compared with the preset confidence P0, and the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor is judged according to the comparison result, and the correction process is adjusted according to the judgment result, wherein:

[0040] When P<P0, it is determined that the correlation between the usage of the user behavior data and the performance bottleneck factor of the virtual reality game is low, and the virtual reality game performance optimization process is not adjusted;

[0041] When P≥P0, it is determined that the usage of the user behavior data is highly correlated with the performance bottleneck factor of the virtual reality game, and the correction process is adjusted, and the adjustment coefficient Q is set, Q=1+(P-P0), and the preset number of user behavior data repetitions J0 is adjusted according to the adjustment coefficient Q, and the adjusted preset number of user behavior data repetitions is set to Jq0, where Jq0=J0×Q.

[0042] Compared with the prior art, the beneficial effect of the present invention is that the method collects virtual reality game performance data through step S1 to fully understand the key indicators of the game such as the running status, resource consumption, frame rate, delay, etc., and provide data support for performance optimization; the method pre-processes the virtual reality game performance data through step S2 to obtain pre-processed data, eliminates noise, errors or inconsistencies in the original data, improves data quality, and thus more accurately reflects the actual situation of game performance, providing a more reliable basis for subsequent analysis and optimization; the method analyzes the virtual reality game performance bottleneck factors according to the pre-processed data through step S3, so as to accurately obtain the key factors that cause the game performance to decline; the method inputs the virtual reality game performance bottleneck factors into the optimization decision model through step S4, and outputs the virtual reality game performance optimization strategy; the performance bottleneck factors are associated with the optimization strategy , recommend corresponding optimization strategies for different bottleneck factors, so as to improve the optimization efficiency of virtual reality game performance. The method judges the usage of user behavior data according to the number of repetitions of user behavior data through step S5, and corrects the virtual reality game performance optimization process according to the judgment result, so as to ensure that the optimization process and the optimization result are closer to the user's needs and expectations. The method judges the correlation between the usage of user behavior data and the virtual reality game performance bottleneck factor according to the association rule mining algorithm through step S6, and adjusts the correction virtual reality game performance optimization process according to the judgment result, so as to reveal the influence mechanism of the usage of user behavior data and the virtual reality game performance bottleneck factor, provide deeper insights for the formulation of optimization strategies, and further improve the effectiveness and pertinence of the optimization strategies by adjusting the optimization process to adapt to the correlation, so as to further improve the optimization efficiency of virtual reality game performance.

[0043] In particular, step S20 fills in missing values ​​of the virtual reality game performance data by interpolation, detects outliers of the virtual reality game performance data by IQR method, retains outliers representing important information, and deletes outliers representing unimportant information, so as to improve data quality and more accurately reflect the actual situation of game performance.

[0044] In particular, the step S21 divides the valid data stored in the form of pixels into image data, and divides the valid data stored in the form of numerical values ​​into data point data according to the storage form of the valid data, so as to improve the efficiency of virtual reality game performance data management and utilization.

[0045] In particular, when the data point data is standardized by the Z-score standardization method in the standardization algorithm, the calculation formula of the standardization algorithm is set to improve data comparability and enhance data analysis accuracy, thereby improving the optimization efficiency of virtual reality game performance.

[0046] In particular, the step S23 gradually extracts features from the image data by using the Canny edge detection method, connects strong edges and weak edges, retains the connected image, and deletes isolated edge points to improve the accuracy of the image data and the adjustability of parameters.

[0047] In particular, step S23 uses principal component analysis to extract features from the standardized data point data, reduces the dimension of the data, reduces the computational cost, and facilitates subsequent data processing and analysis, and inputs the feature image data and the feature data point data into a deep feature fusion model for merging, so as to simultaneously utilize the key information in the two types of data to optimize the virtual reality game performance, thereby further improving the optimization efficiency of the virtual reality game performance.

[0048] In particular, the step S3 inputs the preprocessed data into the bottleneck factor analysis model, and outputs the performance bottleneck factor of the virtual reality game through the bottleneck factor analysis model, so as to accurately locate the performance bottleneck, thereby improving the efficiency of problem solving, avoiding resource waste, and improving overall cost-effectiveness.

[0049] In particular, step S4 constructs an optimization decision model through an expert optimization strategy data set and outputs a virtual reality game performance optimization strategy to achieve targeted optimization, thereby improving the optimization efficiency of virtual reality game performance, and the optimization decision model supports rapid iteration to further improve the optimization efficiency of virtual reality game performance.

[0050] In particular, in step S5, when the number of behavior data repetitions is greater than or equal to the preset number of user behavior data repetitions, the expert optimization strategy data set is updated by pushing the expert optimization strategy data set update window to the virtual reality game device administrator to improve the real-time and accuracy of the optimization strategy and further improve the optimization efficiency of the virtual reality game performance.

[0051] In particular, in step S6, when the confidence between the user behavior data usage and the virtual reality game performance bottleneck factor is greater than the preset confidence, the adjustment coefficient is set to increase as the difference between the confidence and the preset confidence increases, so as to adjust the change in the preset number of repetitions of the user behavior data when the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor is strong, thereby increasing the sensitivity of the game performance optimization process and further improving the optimization efficiency of the virtual reality game performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Schematic diagram of the flow of the game data mining method of this embodiment;

[0053] Figure 2 Schematic diagram of the preprocessing method of this example. DETAILED DESCRIPTION

[0054] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0056] See also Figure 1 As shown, it is a flowchart of the game data mining method of this example, and the method includes:

[0057] Step S1, collecting virtual reality game performance data;

[0058] Step S2, preprocessing the virtual reality game performance data to obtain preprocessed data;

[0059] Step S3, analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data;

[0060] Step S4, inputting the virtual reality game performance bottleneck factor into the optimization decision model, and outputting the virtual reality game performance optimization strategy;

[0061] Step S5, judging the usage of the user behavior data according to the number of repetitions of the user behavior data, and correcting the virtual reality game performance optimization process according to the judgment result;

[0062] Step S6, judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, and adjusting the virtual reality game performance optimization process according to the judgment result.

[0063] Specifically, the method collects virtual reality game performance data through step S1 to fully understand key indicators such as the running status, resource consumption, frame rate, and delay of the game, and provide data support for performance optimization. The method preprocesses the virtual reality game performance data through step S2 to obtain preprocessed data, eliminate noise, errors, or inconsistencies in the original data, and improve data quality, thereby more accurately reflecting the actual situation of game performance and providing a more reliable basis for subsequent analysis and optimization. The method analyzes the virtual reality game performance bottleneck factors according to the preprocessed data through step S3, so as to accurately obtain the key factors that cause the game performance to decline. The method inputs the virtual reality game performance bottleneck factors into the optimization decision model through step S4, and outputs the virtual reality game performance optimization strategy; the performance bottleneck factors are associated with the optimization strategy to provide different bottleneck factors. The method recommends corresponding optimization strategies based on the factors, thereby improving the optimization efficiency of virtual reality game performance. The method judges the usage of user behavior data according to the number of repetitions of user behavior data through step S5, and corrects the virtual reality game performance optimization process according to the judgment result, ensuring that the optimization process and optimization results are closer to user needs and expectations. The method judges the correlation between the usage of user behavior data and the bottleneck factors of virtual reality game performance through step S6 according to the association rule mining algorithm, and adjusts the correction virtual reality game performance optimization process according to the judgment result, revealing the influence mechanism of the usage of user behavior data and the bottleneck factors of virtual reality game performance, providing deeper insights for the formulation of optimization strategies, and further improving the effectiveness and pertinence of the optimization strategies by adjusting the optimization process to adapt to this correlation, thereby further improving the optimization efficiency of virtual reality game performance.

[0064] Specifically, in step S1, when collecting the virtual reality game performance data, the virtual reality game performance data is collected through the internal monitoring system of the virtual reality game device.

[0065] Specifically, the virtual reality game device internal monitoring system refers to a system integrated in the virtual reality game device and game software for real-time capture, analysis and display of various performance data and user behavior data during the game operation process. The virtual reality game refers to a game created using virtual reality technology that allows players to interact and experience in a virtual environment. The virtual reality game performance data refers to a series of quantitative indicators that describe and evaluate the performance of the virtual reality game system. This embodiment does not limit the specific content of the virtual reality game performance data. Relevant technicians in this field can set it freely as long as the needs of subsequent data processing and analysis are met. For example, the virtual reality game performance data can be set to include frame rate, latency, memory usage, processor usage, resolution and perception.

[0066] Specifically, in step S2, the virtual reality game performance data is preprocessed by a preprocessing method to obtain preprocessed virtual reality game performance data.

[0067] See also Figure 2 As shown, it is a schematic diagram of the process of the pretreatment method of this example, and the pretreatment method includes:

[0068] Step S20, cleaning the virtual reality game performance data to obtain valid data;

[0069] Step S21, dividing the valid data into image data and data point data;

[0070] Step S22, standardizing the data point data by using a standardization algorithm to obtain standardized data point data;

[0071] Step S23, extracting features from the image data and the standardized data point data, and merging the feature-extracted data to obtain preprocessed data.

[0072] Specifically, in step S20, when the virtual reality game performance data is cleaned, the missing values ​​of the virtual reality game performance data are filled by the interpolation method, the abnormal values ​​of the virtual reality game performance data are detected by the IQR method, and the representative information of the abnormal values ​​is collected, and the importance of the representative information of the abnormal values ​​is judged. The abnormal values ​​are deleted and marked according to the judgment result to obtain valid data, wherein:

[0073] When the information represented by the outliers is important, keep and mark the outliers;

[0074] When the information represented by the outlier is unimportant, the outlier is deleted.

[0075] Specifically, the interpolation method refers to a numerical calculation method used to construct a function based on given finite data, and the function has the same properties as the original function at these data points. The missing value refers to the clustering, grouping, deletion and truncation of data in the virtual reality game performance data due to lack of information. The IQR method refers to the interquartile range, which is a method of describing data distribution in statistics. The outlier refers to a set of measured values ​​whose deviation from the mean value exceeds a preset multiple, such as a measured value of twice the standard deviation. The representative information of the outlier refers to specific events in the virtual reality game, such as network anomalies, character model loading failure and game rendering failure. The important information refers to information that has a significant impact on the performance of the virtual reality game, such as network anomalies, data packet reading failure and virtual reality game freeze.

[0076] Specifically, in step S21, the valid data stored in the form of pixels is divided into image data, and the valid data stored in the form of numerical values ​​is divided into data point data according to the storage form of the valid data.

[0077] Specifically, the storage form refers to the way in which valid data is represented and saved in a virtual reality system. For example, data can be stored in binary, text, image, audio and video forms. The valid data stored in pixel form refers to data stored in image and video forms. The valid data stored in numerical form refers to data stored in numerical data form.

[0078] Specifically, in step S22, the data point data is standardized by the Z-score standardization method in the standardization algorithm, and the calculation formula of the standardization algorithm is set to Z=(X-μ) / σ, where: X is the data point data, μ is the mean of the data point data, σ is the standard deviation of the data point data, and Z is the standardized data point data;

[0079] The calculation formula for setting the mean μ is Where N is the number of data points in the data point data, Xi is the i-th data point in the data point data, for example, the first data point in the data point data is X1, the second data point is X2, ... the i-th data point is Xi;

[0080] The calculation formula for setting the standard deviation σ is

[0081] Specifically, the standardization algorithm refers to converting data into a distribution with a mean of 0 and a standard deviation of 1, the Z-score standardization method refers to a data preprocessing technique for scaling data of different features to the same scale so that the data of each feature has a standard normal distribution with a mean of 0 and a standard deviation of 1, the mean of the data point data refers to the sum of all data points divided by the number of data points, the standard deviation of the data point data refers to an indicator measuring the degree of dispersion of the data distribution, and the number of data points in the data point data refers to the total number of data points contained in the data point data.

[0082] Specifically, in step S23, feature extraction is performed on the image data by using the Canny edge detection method to obtain feature image data, and the Canny edge detection method includes:

[0083] Step G100, using Gaussian filtering to remove noise from the image data to obtain image data after the noise is removed;

[0084] Step G200, using the Sobel operator to calculate the gradient direction of the image data after the noise is removed, to obtain the gradient direction of the image data;

[0085] Step G300, using a non-maximum suppression method to perform non-maximum suppression in the gradient direction of the image data, retaining local maximum points, and obtaining a maximum point distribution of the image data;

[0086] Step G400, using a double threshold processing method to judge each point in the maximum point distribution of the image data, and performing edge connection based on the judgment result, setting the value of each maximum point to V, the maximum threshold to Vmax, and the minimum threshold to Vmin, where:

[0087] When V>Vmax, the maximum point is determined to be a strong edge;

[0088] When Vmin≤V≤Vmax, the maximum point is determined to be a weak edge;

[0089] When V<Vmin, the maximum point is determined to be a non-edge point;

[0090] Step G500, connecting the strong edge and the weak edge, and outputting the connected image as feature image data.

[0091] Specifically, the Canny edge detection method refers to an image processing technology for extracting edges from an image, the Gaussian filter refers to an image smoothing method that convolves an image with a Gaussian function to reduce noise, the noise removal refers to reducing high-frequency noise in an image through a filter to make the edges and features in the image more obvious, the Sobel operator refers to a convolution kernel used to calculate the gradient of an image, the gradient intensity refers to the amplitude of the change in pixel values ​​in an image, the gradient direction refers to the direction of the change in pixel values ​​in an image, the non-maximum suppression method refers to an edge thinning technology for removing pixels that are not true edges, and the thinned edge refers to suppressing non-maximum points in the gradient direction and retaining only local The local maximum point refers to a pixel point whose gradient value is greater than that of its adjacent points in the gradient direction. The double threshold processing method refers to a method for classifying a gradient image using two thresholds in Canny edge detection. The value of each point refers to the gradient intensity value of each pixel in the gradient image. The maximum threshold refers to the upper limit value used in the double threshold processing method. The minimum threshold refers to the lower limit value used in the double threshold processing method. Connecting the strong edge and the weak edge refers to connecting the strong edge and the weak edge adjacent to it to form a complete edge structure. The isolated weak edge point refers to a weak edge point that is not connected to the strong edge during the double threshold processing and edge connection process.

[0092] Specifically, in step S23, the standardized data point data is also feature extracted by principal component analysis to obtain feature data point data, and the feature image data and feature data point data are input into a deep feature fusion model for merging, and the deep feature fusion model outputs preprocessed data.

[0093] Specifically, the principal component analysis method refers to a statistical technique for converting a set of variables that may be correlated into a set of linearly unrelated variables through orthogonal transformation. The deep feature fusion model refers to a model that meets the preset merging success rate obtained by training the convolutional neural network model according to a preset feature fusion data set. The preset feature fusion data set refers to a data collection of feature image data and feature data point data preset for training the convolutional neural network model. This embodiment does not limit the training method of the convolutional neural network model. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the needs of merging feature image data and feature data point data. For example, it can be set to divide 70% of the preset feature fusion data set into a training set, 20% into a test set, and 10% into a verification set, and input the training set into the convolutional neural network model for training, and input the test set into the trained convolutional neural network model, optimize the parameters in the convolutional neural network model Iteratively, and input the verification set into the convolutional neural network model for verification until the accuracy of the output result of the convolutional neural network model reaches 99%, and the convolutional neural network model is output as a deep feature fusion model.

[0094] Specifically, in step S3, when analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data, the preprocessed data is input into the bottleneck factor analysis model, and the performance bottleneck factors of the virtual reality game are output through the bottleneck factor analysis model.

[0095] Specifically, the virtual reality performance bottleneck factor refers to a factor that causes performance degradation when running virtual reality games and applications, such as hardware performance, software performance, network performance, game content and scene complexity. The bottleneck factor analysis model refers to a model that meets the preset analysis accuracy rate obtained by training the random forest model according to the preset game performance bottleneck factor data set. The preset game performance bottleneck factor data set refers to the preset preprocessing data and the preset virtual reality game performance bottleneck factor corresponding to the preset preprocessing data. This embodiment does not limit the training method of the random forest model. Relevant technicians in this field can freely set it according to actual needs, and only need to meet the needs of analyzing the virtual reality game performance bottleneck factor. For example, it can be set to divide 60% of the preset game performance bottleneck factor data set into a training set, 20% into a test set, and 20% into a validation set, and input the training set into the random forest model for training, and input the test set into the trained random forest model, calibrate the parameters of the random forest model according to the test results, and input the validation set into the random forest model for verification, until the accuracy of the random forest model output result reaches 80%, the random forest model is output as the bottleneck factor analysis model.

[0096] Specifically, in step S4, an optimization decision model is constructed through the expert optimization strategy data set, and the virtual reality game performance bottleneck factor is input into the optimization decision model to output the virtual reality game performance optimization strategy.

[0097] Specifically, the expert optimization strategy data set refers to a strategy set for optimizing virtual reality game performance designed by experts in the field of virtual reality games based on experience and knowledge, including preset virtual reality game performance bottleneck factors and virtual reality game performance optimization strategies corresponding to the preset virtual reality game performance bottleneck factors. The optimization decision model refers to a decision tree model that meets the preset decision accuracy rate and is trained according to the expert optimization strategy data set. This embodiment does not limit the training method of the decision tree model, and relevant technical personnel in this field can freely set it as long as it meets the needs of the optimization decision model. For example, the training method of the decision tree model can be set to divide 90% of the expert optimization strategy data set into a training set and 10% into a verification set, and the maximum depth of the tree in the decision tree model parameter is set to the virtual reality in the expert optimization strategy data set. The number of game performance optimization strategies, the minimum number of samples is set to 1, the training set is input into the decision tree model after setting parameters for training, and the verification set is input into the trained decision tree model, and the parameters of the decision tree model are corrected until the accuracy of the output result of the decision tree model reaches 100%. The maximum depth of the tree refers to the number of nodes in the longest path from the root node to the leaf node in the decision tree model. The minimum number of samples refers to the minimum number of samples that a node needs to contain in the decision tree. The virtual reality game performance optimization strategy refers to improving the efficiency and response speed of the virtual reality game at runtime through a pre-designed method for optimizing the performance of the virtual reality game, such as optimizing the code of the virtual reality game engine, optimizing memory management, using low-polygon models, reducing cutscenes and balancing server loads.

[0098] Specifically, in step S5, the user behavior data is obtained, the number of repetitions J of the user behavior data is compared with the preset number of repetitions J0 of the user behavior data, the use of the user behavior data is judged according to the comparison result, and the compliance of the virtual reality game performance optimization process is judged and corrected according to the judgment result, wherein:

[0099] When J<J0, the user behavior data is determined to be low-frequency user behavior data, the virtual reality game performance optimization process meets the standard, and the virtual reality game performance optimization process is not corrected:

[0100] When J≥J0, the user behavior data is determined to be high-frequency user behavior data, and the virtual reality game performance optimization process is not up to standard. The virtual reality game performance optimization process is corrected, and an expert optimization strategy data set update window is pushed to the virtual reality game device administrator. The virtual reality game device administrator updates the expert optimization strategy data set, obtains the updated expert optimization strategy data set, and optimizes the virtual reality game performance according to the updated expert optimization strategy data set.

[0101] Specifically, the user behavior data refers to data generated by the user when using the virtual reality game device, such as interaction data, motion data, virtual reality game device usage data, user preference data and social interaction data. The user behavior data repetition number J refers to the number of times the user repeatedly performs the same operation in the virtual reality game device. The preset user behavior data repetition number J0 refers to a preset reference value used to reflect the usage of the user behavior data. This embodiment does not limit the value of the preset user behavior data repetition number J0. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the need to judge the usage of the user behavior data. For example, the preset user behavior data repetition number J0 can be set to 20. The virtual reality game device administrator refers to a professional responsible for managing, configuring and maintaining the virtual reality game device. This embodiment does not limit the method of obtaining the user behavior data. Relevant technical personnel in this field can freely set it according to actual needs, and only need to meet the need to obtain the user behavior data. For example, the acquisition method can be set to obtain the user behavior data through the internal monitoring system of the virtual reality game device.

[0102] Specifically, in step S6, when judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, the confidence P of the user behavior data usage and the virtual reality game performance bottleneck factor is calculated according to the association rule mining algorithm, and the confidence P is compared with the preset confidence P0, and the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor is judged according to the comparison result, and the correction process is adjusted according to the judgment result, wherein:

[0103] When P<P0, it is determined that the correlation between the usage of the user behavior data and the performance bottleneck factor of the virtual reality game is low, and the virtual reality game performance optimization process is not adjusted;

[0104] When P≥P0, it is determined that the usage of the user behavior data is highly correlated with the performance bottleneck factor of the virtual reality game, and the correction process is adjusted, and the adjustment coefficient Q is set, Q=1+(P-P0), and the preset number of user behavior data repetitions J0 is adjusted according to the adjustment coefficient Q, and the adjusted preset number of user behavior data repetitions is set to Jq0, where Jq0=J0×Q.

[0105] Specifically, the association rule mining algorithm refers to a data mining algorithm used to discover association relationships between data items from a large-scale data set, the association refers to the relationship and dependency between data items, the confidence refers to an indicator that quantifies the reliability of the association rule, and the preset confidence P0 refers to a preset value used to judge the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor. This embodiment does not limit the value of the preset confidence P0, and relevant technical personnel in this field can freely set it according to actual needs, as long as it meets the need to judge the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor, such as the preset confidence P0 can be set to 0.75.

[0106] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A game data mining method, characterized in that: include: Step S1, collecting virtual reality game performance data; Step S2, preprocessing the virtual reality game performance data by a preprocessing method to obtain preprocessed data; Step S3, analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data; Step S4, inputting the virtual reality game performance bottleneck factor into the optimization decision model, and outputting the virtual reality game performance optimization strategy; Step S5, judging the usage of the user behavior data according to the number of repetitions of the user behavior data, and correcting the virtual reality game performance optimization process according to the judgment result; Step S6, judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, and adjusting the virtual reality game performance optimization process according to the judgment result; In step S5, the user behavior data is obtained, the number of repetitions J of the user behavior data is compared with the preset number of repetitions J0 of the user behavior data, the use of the user behavior data is judged according to the comparison result, and the compliance of the virtual reality game performance optimization process is judged and corrected according to the judgment result, wherein: When J<J0, the user behavior data is determined to be low-frequency user behavior data, the virtual reality game performance optimization process meets the standard, and the virtual reality game performance optimization process is not corrected: When J≥J0, the user behavior data is determined to be high-frequency user behavior data, and the virtual reality game performance optimization process is not up to standard. The virtual reality game performance optimization process is corrected, and an expert optimization strategy data set update window is pushed to the virtual reality game device administrator. The virtual reality game device administrator updates the expert optimization strategy data set, obtains the updated expert optimization strategy data set, and optimizes the virtual reality game performance according to the updated expert optimization strategy data set.

2. The game data mining method according to claim 1, characterized in that: In the step S2, the preprocessing method includes: step S20, performing data cleaning on the virtual reality game performance data to obtain valid data; Step S21, dividing the valid data into image data and data point data; Step S22, standardizing the data point data by using a standardization algorithm to obtain standardized data point data; Step S23, extracting features from the image data and the standardized data point data, and merging the feature-extracted data to obtain preprocessed data.

3. The game data mining method according to claim 2, characterized in that: In step S20, when the virtual reality game performance data is cleaned, the missing values ​​of the virtual reality game performance data are filled by the interpolation method, the abnormal values ​​of the virtual reality game performance data are detected by the IQR method, and the representative information of the abnormal values ​​is collected, and the importance of the representative information of the abnormal values ​​is judged. The abnormal values ​​are deleted and marked according to the judgment result to obtain valid data, wherein: When the information represented by the outliers is important, keep and mark the outliers; When the information represented by the outlier is unimportant, the outlier is deleted.

4. The game data mining method according to claim 3, characterized in that: In the step S21, the valid data stored in the form of pixels is divided into image data, and the valid data stored in the form of numerical values ​​is divided into data point data according to the storage form of the valid data; In the step S22, the data point data is standardized by the Z-score standardization method in the standardization algorithm, and the calculation formula of the standardization algorithm is set to Z=(X-μ) / σ, where: X is the data point data, μ is the mean of the data point data, σ is the standard deviation of the data point data, and Z is the standardized data point data; The calculation formula for setting the mean μ is: where N is the number of data points in the data point data, Xi is the i-th data point in the data point data, the first data point in the data point data is X1, the second data point is X2, ... the i-th data point is Xi; The calculation formula for setting the standard deviation σ is 5. The game data mining method according to claim 4, characterized in that: In the step S23, feature extraction is performed on the image data by using the Canny edge detection method to obtain feature image data, and the Canny edge detection method includes: step G100, using Gaussian filtering to remove noise from the image data to obtain image data after noise removal; Step G200, using the Sobel operator to calculate the gradient direction of the image data after the noise is removed, to obtain the gradient direction of the image data; Step G300, using a non-maximum suppression method to perform non-maximum suppression in the gradient direction of the image data, retaining local maximum points, and obtaining a maximum point distribution of the image data; Step G400, using a double threshold processing method to judge each point in the maximum point distribution of the image data, and performing edge connection according to the judgment result, setting the value of each maximum point to V, the maximum threshold to Vmax, and the minimum threshold to Vmin, wherein: when V>Vmax, the maximum point is determined to be a strong edge; When Vmin≤V≤Vmax, the maximum point is determined to be a weak edge; When V<Vmin, the maximum point is determined to be a non-edge point; Step G500, connecting the strong edge and the weak edge, and outputting the connected image as feature image data.

6. The game data mining method according to claim 5, characterized in that: In step S23, the principal component analysis method is also used to extract features from the standardized data point data to obtain feature data point data, and the feature image data and feature data point data are input into a deep feature fusion model for merging, and the deep feature fusion model outputs preprocessed data.

7. The game data mining method according to claim 1, characterized in that: In step S3, when analyzing the performance bottleneck factors of the virtual reality game based on the preprocessed data, the preprocessed data is input into the bottleneck factor analysis model, and the performance bottleneck factors of the virtual reality game are output through the bottleneck factor analysis model.

8. The game data mining method according to claim 1, characterized in that: In step S4, an optimization decision model is constructed through the expert optimization strategy data set, and the virtual reality game performance bottleneck factor is input into the optimization decision model to output the virtual reality game performance optimization strategy.

9. The game data mining method according to claim 1, characterized in that: In step S6, when judging the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor according to the association rule mining algorithm, the confidence P of the user behavior data usage and the virtual reality game performance bottleneck factor is calculated according to the association rule mining algorithm, and the confidence P is compared with the preset confidence P0, and the correlation between the user behavior data usage and the virtual reality game performance bottleneck factor is judged according to the comparison result, and the correction process is adjusted according to the judgment result, wherein: When P<P0, it is determined that the correlation between the usage of the user behavior data and the performance bottleneck factor of the virtual reality game is low, and the virtual reality game performance optimization process is not adjusted; When P≥P0, it is determined that the usage of the user behavior data is highly correlated with the performance bottleneck factor of the virtual reality game, and the correction process is adjusted, and the adjustment coefficient Q is set, Q=1+(P-P0), and the preset number of user behavior data repetitions J0 is adjusted according to the adjustment coefficient Q, and the adjusted preset number of user behavior data repetitions is set to Jq0, where Jq0=J0×Q.

Images