A game data analysis system based on machine learning

Abstract

The present invention belongs to the field of game data analysis, and discloses a game data analysis system based on machine learning, including an analysis terminal, an authorization terminal and a database terminal; the authorization terminal is used to set multiple database authority ranges for data analysts, including: setting corresponding database ranges for each type of authority verification data respectively; the database terminal is used to store the corresponding database authority ranges for each type of authority verification data, and store game data; the analysis terminal is used to obtain the authority verification data and the download request, and send the authority verification data and the download request to the database terminal; the database terminal is used to determine whether the download request is legal according to the authority verification data, and if so, return the corresponding data to the analysis terminal according to the download request; the analysis terminal is used to analyze the data returned by the database terminal to obtain the analysis result. The present invention effectively improves the efficiency of data analysis.

Description

Technical Field

[0001] The present invention relates to the field of game data analysis, and in particular to a game data analysis system based on machine learning. Background Art

[0002] Game data analysis generally refers to the analysis of data such as the player's online time and recharge amount. By analyzing game data, game activities can be recommended to players more accurately. The game data analysis system generally includes a database for storing game data, an analysis terminal, and an authorization management terminal. The analysis terminal is an electronic device installed with analysis software, and the authorization management terminal is a terminal for managing the database permissions of data analysts using the analysis terminal. When using the analysis terminal, the data analyst can download the game data stored in the database to the analysis terminal for further analysis and processing within the scope of the database permissions he has. In order to ensure the security of game data, the database permission range is fixed. Therefore, if the data analyst needs to use a higher permission to download the game data during analysis, he needs to apply to the authorization manager for a temporary grant of a larger database permission range. This permission management process affects the efficiency of data analysis and is not conducive to data analysts obtaining analysis results more efficiently. Summary of the invention

[0003] The purpose of the present invention is to disclose a game data analysis system based on machine learning to solve the technical problems raised in the background technology.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] The present invention provides a game data analysis system based on machine learning, including an analysis terminal, an authorization terminal and a database terminal;

[0006] The authorization terminal is used to set multiple database permission ranges for data analysts, including:

[0007] Set the corresponding database scope for each type of permission verification data;

[0008] The database terminal is used to store the database authority range corresponding to each type of authority verification data setting, and to store game data;

[0009] The analysis terminal is used to obtain the authority verification data and the download request, and send the authority verification data and the download request to the database terminal;

[0010] The database terminal is used to determine whether the download request is legal based on the authority verification data, and if so, return the corresponding data to the analysis terminal based on the download request;

[0011] The analysis terminal is used to analyze the data returned by the database terminal to obtain analysis results.

[0012] Preferably, the types of authority verification data include a first type, a second type and a third type;

[0013] The first type of authority verification data includes an account number and password;

[0014] The second type of permission verification data includes account password and face image;

[0015] The third type of authority verification data includes account password, face image and fingerprint image.

[0016] Preferably, a corresponding database authority range is set for each type of authority verification data, including:

[0017] For the first type of permission verification data, the corresponding database permission range is set to the 1st column to the N1th column of the table storing the game data;

[0018] For the second type of permission verification data, the corresponding database permission range is set to the 1st column to the N2th column of the table storing the game data;

[0019] For the third type of permission verification data, the corresponding database permission range is set to the 1st column to the N3th column of the table storing the game data;

[0020] N1<N2<N3.

[0021] Preferably, the download request includes the column numbers of the start column and the end column requested to be downloaded in the table storing the game data.

[0022] Preferably, judging whether the download request is legal based on the authority verification data includes:

[0023] If the type of the authority verification data is the first type, then determine whether the account and password sent by the analysis terminal are correct. If so, then determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, then it indicates that the download request is illegal.

[0024] If the type of the authority verification data is the second type, then determine whether the account password and face image sent by the analysis terminal are correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, it indicates that the download request is illegal.

[0025] If the type of authority verification data is the third type, then determine whether the account password, face image and fingerprint image sent by the analysis terminal are correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request; if not, it means that the download request is illegal.

[0026] Preferably, when the type of the authority verification data is the first type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0027] Determine whether the number of the termination column is less than or equal to N1. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal;

[0028] When the type of the authority verification data is the second type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0029] Determine whether the number of the termination column is less than or equal to N2. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal;

[0030] When the type of the authority verification data is the third type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0031] It is determined whether the number of the termination column is less than or equal to N3. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal.

[0032] Preferably, determining whether the account and password sent by the analysis terminal is correct includes:

[0033] The database terminal pre-stores accounts with downloading authority and passwords corresponding to each account;

[0034] Query the password corresponding to the account sent by the analysis terminal from the passwords stored in the database terminal;

[0035] Determine whether the password obtained by the query is consistent with the password sent by the analysis terminal. If so, it means that the account and password sent by the analysis terminal are correct. If not, it means that the account and password sent by the analysis terminal are incorrect.

[0036] Determine whether the facial image sent by the analysis terminal is correct, including:

[0037] The database terminal pre-stores accounts with downloading authority and facial images corresponding to each account;

[0038] Query and analyze the facial image corresponding to the account sent by the terminal from the facial images stored in the database terminal;

[0039] Calculate a first similarity between the face image obtained by the query and the face image sent by the analysis terminal. If the first similarity is greater than a preset first threshold, the face image sent by the analysis terminal is correct; otherwise, the face image sent by the analysis terminal is incorrect.

[0040] Determine whether the fingerprint image sent by the analysis terminal is correct, including:

[0041] The database terminal pre-stores accounts with downloading authority and fingerprint images corresponding to each account;

[0042] Querying the fingerprint image corresponding to the account sent by the analysis terminal from the fingerprint images stored in the database terminal;

[0043] The second similarity between the fingerprint image obtained by the query and the fingerprint image sent by the analysis terminal is calculated. If the second similarity is greater than a preset second threshold, the fingerprint image sent by the analysis terminal is correct, otherwise, the fingerprint image sent by the analysis terminal is incorrect.

[0044] Preferably, returning corresponding data to the analysis terminal according to the download request includes:

[0045] The data of all columns ranging from the start column to the end column in the table storing the game data are sent to the analysis terminal.

[0046] Preferably, analyzing the data returned by the database terminal to obtain analysis results includes:

[0047] Use clustering algorithm to perform clustering calculation on the data returned by the database terminal to obtain analysis results.

[0048] Preferably, the clustering algorithm comprises a K-means clustering algorithm.

[0049] Beneficial effects:

[0050] Compared with the prior art, the database terminal of the present invention performs permission verification on different types of permission verification data provided by data analysts. Therefore, data analysts can obtain a larger permission range by changing the type of permission verification data provided, thereby ensuring the security of game data stored in the database terminal while enabling data analysts to have a more flexible permission range, reducing the probability of the occurrence of events in which the permission range needs to be changed again through the authorization terminal, and effectively improving the efficiency of data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0052] Figure 1 A schematic diagram of a game data analysis system based on machine learning of the present invention.

[0053] Figure 2 Another schematic diagram of a game data analysis system based on machine learning of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] like Figure 1 As shown, the present invention provides a game data analysis system based on machine learning, including an analysis terminal, an authorization terminal and a database terminal;

[0056] The authorization terminal is used to set multiple database permission ranges for data analysts, including:

[0057] Set the corresponding database scope for each type of permission verification data;

[0058] The authorization terminal can send the set database range to the database terminal, and the database terminal is used to store the database authority range corresponding to each type of authority verification data setting, and store game data;

[0059] The analysis terminal is used to obtain the authority verification data and the download request, and send the authority verification data and the download request to the database terminal;

[0060] The database terminal is used to determine whether the download request is legal based on the authority verification data, and if so, return the corresponding data to the analysis terminal based on the download request;

[0061] The analysis terminal is used to analyze the data returned by the database terminal to obtain analysis results.

[0062] The database terminal of the present invention performs permission verification on different types of permission verification data provided by data analysts. Therefore, data analysts can obtain higher permissions by changing the type of permission verification data provided, thereby ensuring the security of game data stored in the database terminal while enabling data analysts to have more flexible permission scopes, reducing the probability of events requiring re-changing permission scopes through the authorization terminal, and effectively improving the efficiency of data analysis.

[0063] Specifically, the authorized terminal is a terminal used by a person with a higher rank than the data analyst, such as a terminal used by the supervisor of the data analyst or a person in charge of data security.

[0064] Specifically, game data includes each game player’s login time, game duration, completed tasks, recharge amount, list of purchased props, etc.

[0065] Preferably, the types of authority verification data include a first type, a second type and a third type;

[0066] The first type of authority verification data includes an account number and password;

[0067] The second type of permission verification data includes account password and face image;

[0068] The third type of authority verification data includes account password, face image and fingerprint image.

[0069] Specifically, the authority verification data of the data analyst is entered by the authorization terminal and sent to the database terminal for storage.

[0070] like Figure 2 As shown, the authorization terminal includes a camera and a fingerprint sensor. The camera is used to collect facial images of data analysts; the fingerprint sensor is used to obtain fingerprint images of data analysts.

[0071] The fingerprint image may be a fingerprint image of an index finger.

[0072] Correspondingly, the analysis terminal also includes a camera and a fingerprint sensor. The camera is used to collect facial images of data analysts; the fingerprint sensor is used to obtain fingerprint images of data analysts.

[0073] Preferably, a corresponding database authority range is set for each type of authority verification data, including:

[0074] For the first type of permission verification data, the corresponding database permission range is set to the 1st column to the N1th column of the table storing the game data;

[0075] For the second type of permission verification data, the corresponding database permission range is set to the 1st column to the N2th column of the table storing the game data;

[0076] For the third type of permission verification data, the corresponding database permission range is set to the 1st column to the N3th column of the table storing the game data;

[0077] N1<N2<N3.

[0078] The authorization terminal of the present invention can set different permission ranges for different types of permission verification data. The more types of data contained in the permission verification data, the larger the permission range. Therefore, when the data analyst needs a larger permission range, he only needs to change the type of permission verification data sent to the database terminal. Thus, different permission ranges can be provided for the same data analyst while ensuring the security of the game data.

[0079] For example, N1 is 10, N2 is 15, and N3 is 20. Data security can be improved by placing more important data in a position where the number of columns of the table storing game data is larger.

[0080] Specifically, each column of the table storing game data corresponds to a type of data, and each row corresponds to a different type of data for the same user.

[0081] Specifically, the first column of the table storing game data is the ID of the game user, and the following columns are login time, recharge amount, etc.

[0082] Preferably, the download request includes the column numbers of the start column and the end column requested to be downloaded in the table storing the game data.

[0083] For example, the starting column is column 3 and the ending column is column 10.

[0084] Preferably, the download request also includes the IP address of the analysis terminal.

[0085] The database terminal sends the data determined according to the download request to the analysis terminal according to the IP address.

[0086] Preferably, judging whether the download request is legal based on the authority verification data includes:

[0087] If the type of the authority verification data is the first type, then determine whether the account and password sent by the analysis terminal are correct. If so, then determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, then it indicates that the download request is illegal.

[0088] If the type of the authority verification data is the second type, then determine whether the account password and face image sent by the analysis terminal are correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, it indicates that the download request is illegal.

[0089] If the type of authority verification data is the third type, then determine whether the account password, face image and fingerprint image sent by the analysis terminal are correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request; if not, it means that the download request is illegal.

[0090] When the database terminal determines whether the download request is legal, it first determines whether the authority verification data is correct, and then determines whether the download request is legal.

[0091] Preferably, when the type of the authority verification data is the first type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0092] Determine whether the number of the termination column is less than or equal to N1. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal;

[0093] When the type of the authority verification data is the second type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0094] Determine whether the number of the termination column is less than or equal to N2. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal;

[0095] When the type of the authority verification data is the third type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows:

[0096] It is determined whether the number of the termination column is less than or equal to N3. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal.

[0097] The corresponding download range is obtained by the type of the permission verification data. After that, the legality of the download request can be determined by judging whether the range in the download request is within the download range. If the range transferred to you by the download request is larger than the download range, the download request is illegal. For example, if the type of the permission verification data is the first type, but the number of columns of the termination column is greater than N1, the download request is illegal.

[0098] Preferably, determining whether the account and password sent by the analysis terminal is correct includes:

[0099] The database terminal pre-stores accounts with downloading authority and passwords corresponding to each account;

[0100] Query the password corresponding to the account sent by the analysis terminal from the passwords stored in the database terminal;

[0101] Determine whether the password obtained by the query is consistent with the password sent by the analysis terminal. If so, it means that the account and password sent by the analysis terminal are correct. If not, it means that the account and password sent by the analysis terminal are incorrect.

[0102] Determine whether the facial image sent by the analysis terminal is correct, including:

[0103] The database terminal pre-stores accounts with downloading authority and facial images corresponding to each account;

[0104] Query and analyze the facial image corresponding to the account sent by the terminal from the facial images stored in the database terminal;

[0105] Calculate a first similarity between the face image obtained by the query and the face image sent by the analysis terminal. If the first similarity is greater than a preset first threshold, the face image sent by the analysis terminal is correct; otherwise, the face image sent by the analysis terminal is incorrect.

[0106] Determine whether the fingerprint image sent by the analysis terminal is correct, including:

[0107] The database terminal pre-stores accounts with downloading authority and fingerprint images corresponding to each account;

[0108] Querying the fingerprint image corresponding to the account sent by the analysis terminal from the fingerprint images stored in the database terminal;

[0109] The second similarity between the fingerprint image obtained by the query and the fingerprint image sent by the analysis terminal is calculated. If the second similarity is greater than a preset second threshold, the fingerprint image sent by the analysis terminal is correct, otherwise, the fingerprint image sent by the analysis terminal is incorrect.

[0110] Specifically, the account and password of the data analyst can be entered by the authorization terminal, and then when the person using the authorization terminal sets the authority scope for the data analyst and enters information such as face images and fingerprint images, the account password, authority scope, face images, and fingerprint images are sent to the database terminal at the same time.

[0111] Preferably, calculating the first similarity between the face image obtained by the query and the face image sent by the analysis terminal includes:

[0112] PA and PB are used to represent the face image obtained by the query and the face image sent by the analysis terminal respectively;

[0113] Use the noise reduction algorithm to reduce the noise of PA and PB respectively to obtain LA and LB;

[0114] Use segmentation algorithm to extract images FA and FB of face regions in LA and LB respectively;

[0115] Calculate the similarity between FA and FB.

[0116] Specifically, performing noise reduction processing before calculating the similarity can effectively reduce the noise content in the image and obtain a more accurate first similarity.

[0117] Preferably, a noise reduction algorithm is used to reduce noise on PA and PB respectively to obtain LA and LB, including:

[0118] For an image p, , the process of denoising p is as follows:

[0119] Get the grayscale image gryp corresponding to p;

[0120] Using the first noise reduction algorithm to reduce the noise of gryp to obtain the image Lp;

[0121] Calculate the grayscale change value of each pixel in Lp respectively;

[0122] The coordinates of the pixel points whose grayscale change value is greater than the grayscale change value threshold are saved to the set LCU;

[0123] Get the image L of the brightness component of p in the Lab color model;

[0124] Use the second noise reduction algorithm to reduce noise on the pixel points whose coordinates belong to LCU in L to obtain image WL;

[0125] WL and Lp are fused to obtain the final denoised image WLP.

[0126] In the process of noise reduction, the present invention does not perform noise reduction processing based on only one noise reduction algorithm, because such noise reduction strength may not be enough, and it is easy to make the grayscale value of the pixel point after noise reduction still differ greatly from the original grayscale value. Therefore, after using the first noise reduction algorithm to perform noise reduction, the present invention determines the coordinates of the pixel points participating in the noise reduction process of another noise reduction algorithm by the grayscale change value, and then uses another noise reduction algorithm in L to perform noise reduction, and finally merges the two different noise reduction algorithms to obtain the final noise reduction result, so as to effectively increase the noise reduction strength and obtain a more effective noise reduction result. When using another noise reduction algorithm to perform noise reduction, the present invention does not perform noise reduction directly on Lp, because this easily causes the edge of the image after noise reduction to be too blurred and loses more edge information. Therefore, the present invention performs noise reduction on the brightness component and then combines the two noise reduction results, so that more edge information can be retained while achieving greater noise reduction.

[0127] In the process of using the second noise reduction algorithm to reduce noise, the present invention does not directly reduce noise for all pixels, because this will affect the efficiency of noise reduction due to the excessive number of pixels that need to reduce noise. The present invention uses grayscale change values ​​for screening, thereby achieving the acquisition of pixels that need to increase noise reduction efforts, and improving the overall noise reduction efficiency.

[0128] Preferably, the calculation formula of the grayscale change value is:

[0129]

[0130] Represents the grayscale change value of pixel b in Lp, Indicates the grayscale value of the pixel in gryp that has the same coordinates as pixel b. Represents the gray value of pixel b in Lp, represents the maximum grayscale value of the pixel in Lp, ub represents the set of pixels in the 8-neighborhood of pixel b in Lp, Represents the image gradient value of the pixel point with the same coordinates as pixel point i in gryp, Represents the image gradient value of pixel i in Lp, Represents the maximum value of the image gradient value of the pixel in Lp, and Represent the grayscale weight and image gradient weight respectively.

[0131] The grayscale change value of the present invention is obtained by comprehensive calculation from two angles: the difference in grayscale values ​​of pixels at the same position in gryp and Lp and the difference in the mean change of neighboring pixels. The greater the change in grayscale value of the pixel before and after denoising, the greater the image gradient of the neighboring pixels before and after denoising. The pixel is assigned a larger grayscale change value, thereby improving the denoising strength for pixels with a high probability of being noise pixels and obtaining better denoising results.

[0132] Preferably, the grayscale change value threshold is 0.8 times the maximum grayscale change value of the pixel points in Lp.

[0133] Specifically, the grayscale weight and image gradient weight are and .

[0134] Preferably, using a second noise reduction algorithm to reduce noise in L for pixels whose coordinates belong to LCU to obtain an image WL includes:

[0135] Get the set PLU of pixels whose coordinates belong to LCU in L;

[0136] Calculate the denoising order value of each pixel in the PLU respectively;

[0137] Each pixel in the PLU is denoised based on the denoising order value to obtain an image WL.

[0138] When using the second noise reduction algorithm for noise reduction, the present invention does not directly perform noise reduction on all pixels on L, but distinguishes the noise reduction order of different pixels, so that the pixels with a later noise reduction order can obtain more effective information, thereby improving the final noise reduction effect.

[0139] Preferably, the calculation formula of the noise reduction order value is:

[0140]

[0141] Indicates the denoising order value of pixel c, Indicates the number of pixels that meet the preset conditions in the 8-neighborhood of pixel c. Indicates the number of pixels in the 8-neighborhood of pixel c that do not belong to the set PLU. To calculate the coefficient, the process of obtaining the pixel points that meet the preset conditions is as follows:

[0142] Use the segmentation algorithm to calculate L and obtain the segmentation threshold thrL;

[0143] For a pixel u in L, if the grayscale value of u is greater than thrL and there is a pixel with a grayscale value less than thrL in the 8-neighborhood of u, then u is a pixel that meets the preset conditions.

[0144] In the present invention, if the number of pixel points that meet the preset conditions around the pixel point is greater and the number of pixel points that do not belong to the set PLU is greater, the denoising order will be higher. In this way, while reducing the influence of the pixel points in the PLU on the denoising result, that is, the influence of the denoising noise on the denoising result, a more accurate denoising result can be obtained by relying on information that meets the preset conditions.

[0145] In the present invention, the pixel points that meet the preset conditions are the pixel points at the intersection of the foreground and the background. These pixel points have boundary information. Therefore, denoising based on the boundary information is conducive to obtaining a more accurate denoising result.

[0146] Specifically, the value range of the calculation coefficient is .

[0147] Preferably, the value of the calculation coefficient is 0.6.

[0148] Preferably, each pixel in the PLU is denoised based on the denoising order value to obtain the image WL, including:

[0149] In the first step, set the value of z to 1, obtain the pixel with the largest denoising order value from PLU as the zth pixel to be denoised, use the second denoising algorithm to denoise the zth pixel to be denoised in L, obtain the grayscale value of the zth pixel to be denoised after denoising, and use the grayscale value of the zth pixel to be denoised after denoising to replace the grayscale value of the zth pixel to be denoised in L to obtain the image ;

[0150] The second step is to delete the zth pixel to be denoised from the PLU;

[0151] The third step is to add 1 to the value of z;

[0152] The fourth step is to obtain the pixel with the largest denoising order value from the PLU as the zth pixel to be denoised, and use the second denoising algorithm to In the process, the zth pixel to be denoised is denoised to obtain the grayscale value of the zth pixel to be denoised after denoising, and the grayscale value of the zth pixel to be denoised after denoising is used to replace The gray value of the zth pixel to be denoised in the image is obtained ;

[0153] Step 5: Delete the zth pixel to be denoised from the PLU;

[0154] Step 6: Determine whether the number of pixels in the PLU is greater than 0. If so, proceed to step 3. If not, As image WL.

[0155] In the denoising process of the present invention, the denoising of the latter pixel is performed on the basis of the image generated by the grayscale value of the previous pixel after denoising. Therefore, the grayscale value of the pixel that has been denoised can effectively participate in the denoising process of the un-denoised pixel, making the grayscale value that can be referenced in the denoising process of the un-denoised pixel more accurate, thereby improving the quality of the final denoising result.

[0156] Preferably, the first denoising algorithm comprises a wavelet denoising algorithm.

[0157] Specifically, the first noise reduction algorithm may also be a Gaussian filtering algorithm or the like.

[0158] Preferably, the second noise reduction algorithm comprises a bilateral filtering algorithm.

[0159] Preferably, WL and Lp are fused to obtain the final denoised image WLP, including:

[0160] In Lp, the grayscale value of each pixel in LCU after fusion is calculated respectively, and the original grayscale value is replaced by the fused grayscale value to obtain the image WLP;

[0161] For a pixel v in an LCU, the formula for calculating the grayscale value of v after fusion is:

[0162]

[0163] Represents the gray value after v fusion, Indicates the gray value of pixel v in Lp, Indicates the grayscale value of the pixel with the same coordinate as v in WL.

[0164] By fusing the results of noise reduction, a stronger noise reduction can be achieved.

[0165] Preferably, a segmentation algorithm is used to extract images FA and FB of the face regions in LA and LB, respectively, including:

[0166] Use the segmentation algorithm to calculate LA and LB respectively to obtain the segmentation thresholds thrA and thrB;

[0167] The pixels in LA whose grayscale values ​​are greater than thrA are taken as the pixels in FA; the pixels in LB whose grayscale values ​​are greater than thrB are taken as the pixels in FB.

[0168] By extracting the face area to calculate the similarity, the influence of background factors on the similarity calculation results can be more effectively reduced.

[0169] Preferably, the segmentation algorithm comprises a maximum inter-class variance method.

[0170] Preferably, the preset first threshold is 0.7.

[0171] Preferably, the preset second threshold is 0.8.

[0172] Preferably, returning corresponding data to the analysis terminal according to the download request includes:

[0173] The data of all columns ranging from the start column to the end column in the table storing the game data are sent to the analysis terminal.

[0174] For example, when the starting column is the 3rd column and the ending column is the 5th column, data of all columns ranging from the 3rd column to the 5th column in the table storing the game data are sent to the analysis terminal.

[0175] Preferably, analyzing the data returned by the database terminal to obtain analysis results includes:

[0176] Use clustering algorithm to perform clustering calculation on the data returned by the database terminal to obtain analysis results.

[0177] Specifically, a clustering algorithm is used to perform clustering calculations on the data returned by the database terminal to obtain analysis results, including:

[0178] A clustering algorithm is used to cluster the data returned by the database terminal to obtain a set of different types of player groups, and the obtained set is used as the analysis result.

[0179] For example, types of player groups include:

[0180] Casual players: They spend less time playing games and prefer simpler game content.

[0181] Core players: They invest more time and pursue the depth and complexity of the game.

[0182] Paying players: tend to make in-app purchases and are willing to pay for additional content in the game.

[0183] Non-paying players: mainly rely on free content and rarely pay.

[0184] Specifically, the clustering process is as follows:

[0185] Step 1: Data preprocessing.

[0186] Each type of data is standardized separately to ensure consistent ranges for different features.

[0187] Step 2: Determine the clustering algorithm.

[0188] Common clustering algorithms include K-means clustering, hierarchical clustering, DBSCAN, etc.

[0189] Step 3: Determine the number of clusters.

[0190] If you choose K-means clustering, you need to determine the number of clusters. You can use the Elbow Method or Silhouette Coefficient to help determine the number of clusters.

[0191] Step 4: Apply clustering algorithm.

[0192] Use the determined clustering algorithm to perform clustering operations on the prepared data.

[0193] In another embodiment, analyzing the data returned by the database terminal to obtain the analysis result includes:

[0194] Use the preset statistical strategy to count the data returned by the database terminal, and use the statistical results as analysis results.

[0195] For example, statistics on the average recharge amount and average daily online time of game users.

[0196] Preferably, the clustering algorithm comprises a K-means clustering algorithm.

[0197] Specifically, the clustering algorithm may also be hierarchical clustering, DBSCAN, etc.

[0198] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A game data analysis system based on machine learning, characterized in that: Including analysis terminal, authorization terminal and database terminal; The authorization terminal is used to set multiple database permission ranges for data analysts, including: Set the corresponding database scope for each type of permission verification data; The database terminal is used to store the database authority range corresponding to each type of authority verification data setting, and to store game data; The analysis terminal is used to obtain the authority verification data and the download request, and send the authority verification data and the download request to the database terminal; The database terminal is used to determine whether the download request is legal based on the authority verification data, and if so, return the corresponding data to the analysis terminal based on the download request; The analysis terminal is used to analyze the data returned by the database terminal and obtain the analysis results; The types of authority verification data include the first type, the second type and the third type; The first type of authority verification data includes an account number and password; The second type of permission verification data includes account password and face image; The third type of permission verification data includes account password, face image and fingerprint image; Set the corresponding database permission range for each type of permission verification data, including: For the first type of permission verification data, the corresponding database permission range is set to the 1st column to the N1th column of the table storing the game data; For the second type of permission verification data, the corresponding database permission range is set to the 1st column to the N2th column of the table storing the game data; For the third type of permission verification data, the corresponding database permission range is set to the 1st column to the N3th column of the table storing the game data; N1<N2<N3; The download request includes the column numbers of the start column and the end column requested to be downloaded in the table storing the game data; Determine whether the download request is legitimate based on the permission verification data, including: If the type of the authority verification data is the first type, then determine whether the account and password sent by the analysis terminal are correct. If so, then determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, then it indicates that the download request is illegal. If the type of the authority verification data is the second type, then determine whether the account password and face image sent by the analysis terminal are correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request. If not, it indicates that the download request is illegal. If the type of the authority verification data is the third type, then determine whether the account password, face image and fingerprint image sent by the analysis terminal are all correct. If so, determine whether the download request is legal based on the number of columns in the termination column in the download request; if not, it indicates that the download request is illegal; Determine whether the account and password sent by the analysis terminal are correct, including: The database terminal pre-stores accounts with downloading authority and passwords corresponding to each account; Query the password corresponding to the account sent by the analysis terminal from the passwords stored in the database terminal; Determine whether the password obtained by the query is consistent with the password sent by the analysis terminal. If so, it means that the account and password sent by the analysis terminal are correct. If not, it means that the account and password sent by the analysis terminal are incorrect. Determine whether the facial image sent by the analysis terminal is correct, including: The database terminal pre-stores accounts with downloading authority and facial images corresponding to each account; Query and analyze the facial image corresponding to the account sent by the terminal from the facial images stored in the database terminal; Calculate a first similarity between the face image obtained by the query and the face image sent by the analysis terminal. If the first similarity is greater than a preset first threshold, the face image sent by the analysis terminal is correct; otherwise, the face image sent by the analysis terminal is incorrect. Determine whether the fingerprint image sent by the analysis terminal is correct, including: The database terminal pre-stores accounts with downloading authority and fingerprint images corresponding to each account; Querying the fingerprint image corresponding to the account sent by the analysis terminal from the fingerprint images stored in the database terminal; Calculate the second similarity between the fingerprint image obtained by the query and the fingerprint image sent by the analysis terminal. If the second similarity is greater than a preset second threshold, the fingerprint image sent by the analysis terminal is correct; otherwise, the fingerprint image sent by the analysis terminal is incorrect. Calculating the first similarity between the face image obtained by the query and the face image sent by the analysis terminal, including: PA and PB are used to represent the face image obtained by the query and the face image sent by the analysis terminal respectively; Use the noise reduction algorithm to reduce the noise of PA and PB respectively to obtain LA and LB; Use segmentation algorithm to extract images FA and FB of face regions in LA and LB respectively; Calculate the similarity between FA and FB; Use the noise reduction algorithm to reduce the noise of PA and PB respectively to obtain LA and LB, including: For an image p, , the process of denoising p is as follows: Get the grayscale image gryp corresponding to p; Using the first noise reduction algorithm to reduce the noise of gryp to obtain the image Lp; Calculate the grayscale change value of each pixel in Lp respectively; The coordinates of the pixel points whose grayscale change value is greater than the grayscale change value threshold are saved to the set LCU; Get the image L of the brightness component of p in the Lab color model; Use the second noise reduction algorithm to reduce noise on the pixel points whose coordinates belong to LCU in L to obtain image WL; WL and Lp are fused to obtain the final denoised image WLP; The calculation formula of grayscale change value is: Represents the grayscale change value of pixel b in Lp, Indicates the grayscale value of the pixel in gryp that has the same coordinates as pixel b. Represents the gray value of pixel b in Lp, represents the maximum grayscale value of the pixel in Lp, ub represents the set of pixels in the 8-neighborhood of pixel b in Lp, Represents the image gradient value of the pixel point with the same coordinates as pixel point i in gryp, Represents the image gradient value of pixel i in Lp, Represents the maximum value of the image gradient value of the pixel in Lp, and Represent the grayscale weight and image gradient weight respectively.

2. A game data analysis system based on machine learning according to claim 1, characterized in that: When the type of the authority verification data is the first type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows: Determine whether the number of the termination column is less than or equal to N1. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal; When the type of the authority verification data is the second type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows: Determine whether the number of the termination column is less than or equal to N2. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal; When the type of the authority verification data is the third type, the process of judging whether the download request is legal based on the number of columns of the termination column in the download request is as follows: It is determined whether the number of the termination column is less than or equal to N3. If so, it indicates that the download request is legal; if not, it indicates that the download request is illegal.

3. The game data analysis system based on machine learning according to claim 1, characterized in that: According to the download request, the corresponding data is returned to the analysis terminal, including: The data of all columns ranging from the start column to the end column in the table storing the game data are sent to the analysis terminal.

4. The game data analysis system based on machine learning according to claim 1, characterized in that: Analyze the data returned by the database terminal and obtain analysis results, including: Use clustering algorithm to perform clustering calculation on the data returned by the database terminal to obtain analysis results.

5. A game data analysis system based on machine learning according to claim 4, characterized in that: The clustering algorithm includes the K-means clustering algorithm.

Images