Account management method and system based on big data

Abstract

The present invention relates to an account management method and system based on big data, the method comprising: at every preset time period, counting the behavior information of all users within the time period, obtaining the user's location information, and constructing a behavior matrix according to the location information; arranging the behavior matrix according to time sequence, extracting all matrix data of the same account label based on the values ​​at the row and column positions, fitting the matrix data, and creating a behavior curve; clustering the accounts by time period based on the behavior curve to obtain similar accounts in each time period; the present invention counts the behavior information of all accounts at each moment in the form of a matrix, selects the matrix data of the same account in the behavior matrix at different moments, connects the matrix data and fits a curve, which can be used as an account feature; and then abstracts the complex behavior data into a curve, compares the curves, and classifies similar accounts into one category, the operation process is extremely convenient, and there is a lot of room for secondary development.

Description

Technical Field

[0001] The present invention relates to the technical field of account management, and in particular to an account management method and system based on big data. Background Art

[0002] With the popularization of computer equipment, everyone can obtain a large number of convenient services with the help of computer equipment. Of course, they can also provide services to others. The party providing services is often called the developer.

[0003] When developers provide services to users, they often need to manage user accounts. Account management involves using big data technologies and tools to collect, store, analyze and manage a large amount of account-related data. Static data such as basic information, account creation time and account status can be managed with the help of database access operations, but the management process for users' dynamic behavior data is more complicated. How to manage users' dynamic behavior data is the technical problem that the technical solution of the present invention aims to solve. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention proposes an account management method and system based on big data.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for account management based on big data, the method comprising the following steps:

[0007] In the pre-processing stage, all account behaviors are counted and numbered; each account behavior corresponds to a unique and definite number;

[0008] Obtain the user's behavior information in real time according to the permissions granted by the user, and query the corresponding number;

[0009] At every preset time, the behavior information of all users within the time is counted, the location information of the users is obtained, and a behavior matrix is ​​constructed according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information;

[0010] Arrange the behavior matrix in chronological order, extract all matrix data of the same account label based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve;

[0011] Based on the behavior curve, the accounts are clustered by time period to obtain similar accounts in each time period.

[0012] As a further technical solution of the present invention, the steps of counting the behavior information of all users within a preset time period, obtaining the user's location information, and constructing a behavior matrix according to the location information include:

[0013] Obtain the number of access accounts in real time, arrange the number of accounts according to the time axis, and fit the number change function;

[0014] Perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function;

[0015] Intercept the quantity change function within a period, segment the quantity change function within a period, and determine the statistical nodes;

[0016] If the current moment reaches the statistical node, the behavior information of all users in the previous time period is counted, the user's location information is obtained, and a behavior matrix is ​​constructed according to the location information.

[0017] As a further technical solution of the present invention, if the current moment reaches the statistical node, the behavior information of all users in the previous time period is counted, the user's location information is obtained, and the step of constructing a behavior matrix according to the location information includes:

[0018] Obtain the previous statistical node of the current statistical node, and use the time period from the previous statistical node to the current statistical node as the time period to be counted;

[0019] For any account within the statistical time period, obtain behavior information in real time and query the number corresponding to the behavior information;

[0020] Insert the account tag inside the number as the element value;

[0021] Acquire the location information of the account, determine the row and column position according to the location information, and create an element according to the row and column position and the element value;

[0022] Count the elements of all accounts and get the behavior matrix.

[0023] As a further technical solution of the present invention, the steps of arranging the behavior matrix in chronological order, extracting all matrix data of the same account label based on the values ​​in the row and column positions, fitting the matrix data, and creating a behavior curve include:

[0024] Arrange the behavior matrix in chronological order;

[0025] Select the row and column positions in the bottom-level behavior matrix in turn, and query the account labels in the values ​​at the row and column positions;

[0026] Locate elements in other behavior matrices based on account labels, and use the located element positions and element values ​​as matrix data;

[0027] Fit the matrix data to obtain the behavior curve corresponding to the account label.

[0028] As a further technical solution of the present invention, the step of clustering the accounts by time period based on the behavior curve to obtain similar accounts in each time period includes:

[0029] Pair account tags in pairs, and read the behavior curve of any two paired accounts based on the account tags;

[0030] Place the behavior curves on the same horizontal axis and calculate the curve distance;

[0031] All accounts are clustered according to the curve distance to obtain similar accounts within the current time period.

[0032] As a further technical solution of the present invention, the generation process of the behavior curve is:

[0033] Randomly select four matrix data and draw Bezier curves based on the four matrix data; wherein the number of random selections is a preset value;

[0034] Calculate the mean distance between any Bezier curve and other Bezier curves, and select the Bezier curve corresponding to the minimum distance mean as the final behavior curve;

[0035] The Bezier curve is:

[0036] B(α)=(1-α) 3 P0+3(1―α) 2 αP1+3(1―α)α 2 P2+α 3 P3; where P0, P1, P2 and P3 are four selected matrix data, α is a preset parameter, α∈[0,1];

[0037] The calculation process of the curve distance is:

[0038] In the formula, f(x) is the function expression of one of the curves based on the same coordinate axis, g(x) is the function expression of the other curve based on the same coordinate axis; x0 is the common left endpoint of the two curves, and x1 is the right endpoint of the curve with a shorter horizontal axis distance on the same coordinate axis.

[0039] The technical solution of the present invention also provides an account management system based on big data, the system comprising:

[0040] The behavior numbering module is used to count all account behaviors in the preprocessing stage and number the account behaviors; each account behavior corresponds to a unique and definite number;

[0041] The number query module is used to obtain the user's behavior information in real time according to the permissions granted by the user and query the corresponding number;

[0042] A behavior matrix construction module is used to count the behavior information of all users within a preset time period, obtain the user's location information, and construct a behavior matrix according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information;

[0043] The curve fitting module is used to arrange the behavior matrix in chronological order, extract all matrix data of the same account label based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve;

[0044] The account clustering module is used to cluster the accounts by time period based on the behavior curve to obtain similar accounts in each time period.

[0045] As a further technical solution of the present invention, the behavior matrix building module includes:

[0046] Function fitting unit, used to obtain the number of access accounts in real time, arrange the number of accounts according to the time axis, and fit the number change function;

[0047] A period extraction unit is used to perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function;

[0048] A function segmentation unit is used to intercept the quantity change function within a period, segment the quantity change function within a period, and determine the statistical nodes;

[0049] An execution unit is constructed to count the behavior information of all users in the previous time period if the current moment reaches the statistical node, obtain the user's location information, and construct a behavior matrix according to the location information.

[0050] As a further technical solution of the present invention, the curve fitting module includes:

[0051] A matrix arrangement unit, used to arrange the behavior matrix according to the time sequence;

[0052] A label query unit is used to select row and column positions in the bottom-level behavior matrix in sequence and query the account labels in the values ​​at the row and column positions;

[0053] A data locating unit is used to locate elements in other behavior matrices in sequence based on account labels, and use the located element positions and element values ​​as matrix data;

[0054] The fitting execution unit is used to fit the matrix data to obtain the behavior curve corresponding to the account label.

[0055] As a further technical solution of the present invention, the account clustering module includes:

[0056] An account pairing unit is used to pair account tags in pairs and read the behavior curve of any two paired accounts according to the account tags;

[0057] A distance calculation unit, used to place the behavior curves in the same horizontal axis and calculate the curve distance;

[0058] The clustering execution unit is used to cluster all accounts according to the curve distance to obtain similar accounts within the current time period.

[0059] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention counts the behavior information of all accounts at each moment in the form of a matrix, uses the location of the behavior information as the matrix coordinates, and uses the number of the behavior information as the matrix value to obtain the behavior matrix, selects the matrix data of the same account in the behavior matrix at different moments, connects the matrix data and fits a curve, which can be used as the account feature; then, the complex behavior data is abstracted into a curve, the curves are compared, and the same type of accounts are classified into one category, the operation process is extremely convenient, and there is a lot of room for secondary development. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a flowchart of an account management method based on big data.

[0061] Figure 2 A flowchart of the steps for building a behavior matrix in a big data-based account management method.

[0062] Figure 3 A flowchart of the steps for creating a behavior curve in a big data-based account management method.

[0063] Figure 4 The present invention is a flowchart of the steps of clustering accounts in an account management method based on big data.

[0064] Figure 5 This is a structural diagram of an account management system based on big data. DETAILED DESCRIPTION

[0065] The technical solution of this patent is further described in detail below in conjunction with specific implementation methods.

[0066] like Figure 1 As shown, an embodiment of the present invention provides an account management method based on big data, and the method includes the following steps:

[0067] Step S100: In the pre-processing stage, all account behaviors are counted and numbered; each account behavior corresponds to a unique and determined number;

[0068] Regardless of the platform, the types of account behaviors are limited, and there are only a few simple user behaviors. For example, in some information retrieval platforms, user behaviors are only clicking certain buttons or entering text in text boxes, and each account behavior corresponds to a number; in fact, even in some complex systems, such as the Windows operating system, each operation corresponds to a number, and the operations that users can perform are also preset and limited. Each operation behavior corresponds to a unique and definite number.

[0069] Step S200: obtaining the user's behavior information in real time according to the authority granted by the user, and querying the corresponding number;

[0070] The purpose of this application is to monitor the user's operating behavior in the platform, and then manage their accounts, and then provide some services. This requires obtaining the permissions granted by the user in advance. If the user does not grant permissions, the user's behavior information cannot be obtained. It should be noted that when the user uses the platform, it is assumed that the permission is granted, because as a service provider, the user's behavior information must be obtained and stored (at least cached). The information obtained is limited to the scope of the services provided by the platform. The user's behavior information is obtained in real time according to the permissions granted by the user, and the corresponding number can be queried.

[0071] Step S300: At every preset time period, the behavior information of all users within the time period is counted, the location information of the users is obtained, and a behavior matrix is ​​constructed according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information;

[0072] At preset time intervals, that is, a period of time, the behavior information of all users within the time period is counted. The more users there are, the more behavior information is obtained. Within the time period, for any user, the user's location information is obtained to determine a row and column position. By counting all the behavior information of a user and combining it with the user's account label, a data (number set and account label) can be obtained as the value of the row and column position. The row and column positions of all users and their row and column values ​​are counted to obtain a behavior matrix.

[0073] Step S400: Arrange the behavior matrix in chronological order, extract all matrix data of the same account tag based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve;

[0074] By arranging the behavior matrix in chronological order on another dimension (equivalent to generating a three-dimensional matrix), all the behavior information of all users within that period of time can be obtained; in each behavior matrix, all the elements of the same account label (corresponding to the same user) are connected to fit a curve, which is called a behavior curve; it should be noted that the matrix data is actually three-dimensional data, including row and column positions and element values; for better operation, a conversion model can be set in advance to convert the element values ​​into numerical values.

[0075] Step S500: clustering the accounts by time period based on the behavior curve to obtain similar accounts in each time period;

[0076] By taking the behavior curve as the account feature of each user and clustering each account, we can find out which users are of the same type within that period of time.

[0077] In the technical solution of the present invention, after obtaining the same type of users, the method has been executed. Regarding the subsequent application stage, this application does not limit it. On the basis of clustering the users, many management solutions can be implemented. The simplest application is to conduct risk assessment on the account.

[0078] Since there is a clustering result in each time period, it indicates which other accounts any account is similar to in each time period. For the same user, the operation method will not change much (the operation habits will not change much), and the clustering result in each time period will not change much. The accounts of each user are obtained and which users' accounts are classified into the same category. If the classification result suddenly changes, it can be determined that the user operating the account has changed. This is very likely to be an account leak, that is, the user using the account has changed.

[0079] Figure 2 The present invention is a flowchart of the steps of constructing a behavior matrix in an account management method based on big data. The steps of counting the behavior information of all users within a preset time period, obtaining the user's location information, and constructing the behavior matrix according to the location information include:

[0080] Step S301: obtaining the number of access accounts in real time, arranging the number of accounts according to the time axis, and fitting the number change function;

[0081] Step S302: Perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function;

[0082] Step S303: intercepting a quantity change function within a period, segmenting the quantity change function within a period, and determining statistical nodes;

[0083] Step S304: if the current moment reaches the statistical node, the behavior information of all users in the previous time period is counted, the user's location information is obtained, and a behavior matrix is ​​constructed according to the location information.

[0084] The above content specifically limits the construction process of the behavior matrix. For the platform that executes this method, there is at least one access monitoring port for obtaining the current number of access accounts in real time. By arranging the number of accounts in chronological order, a quantity change function can be fitted. The periodicity of the quantity change function can be identified to obtain the period of the quantity change function. In actual situations, the number of users on any platform has obvious periodicity. For example, the number of visits to the working platform is relatively large during working hours, and the number of visits is almost zero during non-working hours. Therefore, the Fourier transform scheme is adopted to select the main period as the period of the quantity change function.

[0085] After the period is determined, the quantity change function within a period is intercepted (generally the previous one or several periods are selected). According to the vertical scale value of the quantity change function, the quantity change function can be segmented. The segmentation process can use existing mathematical analysis methods. When the difference between two function values ​​is less than the preset threshold, they are classified into one section.

[0086] After the segmentation is completed, each segment corresponds to a time length, and the data processing process mentioned in this application is to process all data within a time length.

[0087] Furthermore, if the current moment reaches the statistical node, the behavior information of all users in the previous time period is counted, the location information of the users is obtained, and the steps of constructing a behavior matrix according to the location information include:

[0088] Obtain the previous statistical node of the current statistical node, and use the time period from the previous statistical node to the current statistical node as the time period to be counted;

[0089] For any account within the statistical time period, obtain behavior information in real time and query the number corresponding to the behavior information;

[0090] Insert the account tag inside the number as the element value;

[0091] Acquire the location information of the account, determine the row and column position according to the location information, and create an element according to the row and column position and the element value;

[0092] Count the elements of all accounts and get the behavior matrix.

[0093] The previous statistical node of the current statistical node is obtained, and the time period from the previous statistical node to the current statistical node is used as the time period to be counted, which is called duration.

[0094] For any account within the statistical time period, obtain behavior information in real time, query the number corresponding to the behavior information, insert the account label into the number as the element value; for the convenience of analysis, the number and account label can be input into the preset conversion model to obtain a single numerical value, called the element value.

[0095] On this basis, the location information of the account is obtained, and the row and column positions are determined according to the location information. A matrix element can be obtained based on the row and column positions and their element values. By counting all the matrix elements, a matrix called a behavior matrix can be obtained.

[0096] Figure 3 The flowchart is a step of creating a behavior curve in an account management method based on big data. The steps of arranging the behavior matrix in chronological order, extracting all matrix data of the same account label based on the values ​​in the row and column positions, fitting the matrix data, and creating the behavior curve include:

[0097] Step S401: Arrange the behavior matrix according to time sequence;

[0098] Step S402: Select the row and column positions in the bottom-level behavior matrix in turn, and query the account labels in the values ​​at the row and column positions;

[0099] Step S403: Locate elements in other behavior matrices in turn based on the account labels, and use the located element positions and element values ​​as matrix data;

[0100] Step S404: Fit the matrix data to obtain a behavior curve corresponding to the account label.

[0101] In an example of the technical solution of the present invention, the behavior matrix is ​​arranged in chronological order to obtain a three-dimensional matrix; the bottom-level behavior matrix (the behavior matrix at the beginning of the duration) is queried in another dimension, the bottom-level behavior matrix is ​​traversed, an account is first determined, the account tag of the account is queried, and the elements with the same account tag are located in other behavior matrices according to the account tag, all the located elements are counted, and multiple points are selected in the three-dimensional matrix, which can represent the behavior characteristics of a user in the duration; these points are fitted to obtain a behavior curve, and the behavior curve is used as a behavior feature for a subsequent clustering process.

[0102] Figure 4The present invention is a flowchart of a step of clustering accounts in an account management method based on big data, wherein the step of clustering accounts by time periods based on the behavior curve to obtain similar accounts in each time period includes:

[0103] Step S501: pairing account tags in pairs, and reading behavior curves for any two paired accounts according to the account tags;

[0104] Step S502: placing the behavior curves on the same horizontal axis and calculating the curve distance;

[0105] Step S503: clustering all accounts according to the curve distance to obtain similar accounts within the current time period.

[0106] In existing clustering schemes, whether it is K-means clustering or spectral clustering, or other clustering schemes, distance is a necessary parameter. This application makes specific limitations on the distance, calculates the curve distance between two accounts, and clusters all accounts according to the distance, so that similar users in the current time period can be classified into one category.

[0107] Specifically, the generation process of the behavior curve is:

[0108] Randomly select four matrix data and draw Bezier curves based on the four matrix data; wherein the number of random selections is a preset value;

[0109] Calculate the mean distance between any Bezier curve and other Bezier curves, and select the Bezier curve corresponding to the minimum distance mean as the final behavior curve;

[0110] The Bezier curve is:

[0111] B(α)=(1-α) 3 P0+3(1―α) 2 αP1+3(1―α)α 2 P2+α 3 P3; wherein, P0, P1, P2 and P3 are four selected matrix data, α is a preset parameter, α∈[0,1].

[0112] The above content provides a specific curve fitting process. This application uses a cubic Bezier curve for point fitting. In fact, the higher the degree of the Bezier curve used, the more accurate the fitting. The cubic Bezier curve used in this application is a convenient calculation scheme. By selecting four points in the arranged behavior matrix, a curve can be fitted; for the same arranged behavior matrix, multiple curves can be fitted, and the distance between each fitted curve and all other curves is calculated, the mean distance is calculated, and the curve with the smallest mean distance is selected. A curve with a high similarity to all fitting processes can be obtained as the final behavior curve.

[0113] From the above content, it can be seen that the technical solution of this application also involves the comparison process of curves. Whether it is the clustering stage or the screening stage of the above-mentioned behavior curve, it is necessary to calculate the curve distance. The calculation process provided by this application is as follows:

[0114] The calculation process of the curve distance is:

[0115] In the formula, f(x) is the function expression of one of the curves based on the same coordinate axis, g(x) is the function expression of the other curve based on the same coordinate axis; x0 is the common left endpoint of the two curves, and x1 is the right endpoint of the curve with a shorter horizontal axis distance on the same coordinate axis.

[0116] Place the two curves on the same coordinate axis and align the left endpoints, then calculate the integral of the curve difference. The larger the integral, the greater the difference between the two curves. Therefore, the integral can be directly used as the curve distance. Of course, it can also be multiplied by a preset correction factor.

[0117] It should be noted that the part for calculating the integral needs to be the part where the two curves relatively overlap on the same coordinate axis. Therefore, x0 is the common left endpoint of the two curves, and x1 is the right endpoint of the curve with a shorter horizontal axis distance on the same coordinate axis, thereby ensuring that both curves have values ​​during the integral calculation process.

[0118] Figure 5 1 is a structural diagram of an account management system based on big data. Another object of the present invention is to provide an account management system based on big data. The system 10 includes:

[0119] The behavior numbering module 11 is used to count all account behaviors in the preprocessing stage and number the account behaviors; each account behavior corresponds to a unique and definite number;

[0120] The number query module 12 is used to obtain the user's behavior information in real time according to the authority granted by the user and query the corresponding number;

[0121] The behavior matrix construction module 13 is used to count the behavior information of all users within a preset time period, obtain the user's location information, and construct a behavior matrix according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information;

[0122] The curve fitting module 14 is used to arrange the behavior matrix according to the time sequence, extract all matrix data of the same account label based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve;

[0123] The account clustering module 15 is used to cluster the accounts by time period based on the behavior curve to obtain similar accounts in each time period.

[0124] Furthermore, the behavior matrix construction module 13 includes:

[0125] Function fitting unit, used to obtain the number of access accounts in real time, arrange the number of accounts according to the time axis, and fit the number change function;

[0126] A period extraction unit is used to perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function;

[0127] A function segmentation unit is used to intercept the quantity change function within a period, segment the quantity change function within a period, and determine the statistical nodes;

[0128] An execution unit is constructed to count the behavior information of all users in the previous time period if the current moment reaches the statistical node, obtain the user's location information, and construct a behavior matrix according to the location information.

[0129] Specifically, the curve fitting module 14 includes:

[0130] A matrix arrangement unit, used to arrange the behavior matrix according to the time sequence;

[0131] A label query unit is used to select row and column positions in the bottom-level behavior matrix in sequence and query the account labels in the values ​​at the row and column positions;

[0132] A data locating unit is used to locate elements in other behavior matrices in sequence based on account labels, and use the located element positions and element values ​​as matrix data;

[0133] The fitting execution unit is used to fit the matrix data to obtain the behavior curve corresponding to the account label.

[0134] Furthermore, the account clustering module 15 includes:

[0135] An account pairing unit is used to pair account tags in pairs and read the behavior curve of any two paired accounts according to the account tags;

[0136] A distance calculation unit, used to place the behavior curves in the same horizontal axis and calculate the curve distance;

[0137] The clustering execution unit is used to cluster all accounts according to the curve distance to obtain similar accounts within the current time period.

[0138] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An account management method based on big data, characterized in that: The method comprises the following steps: In the pre-processing stage, all account behaviors are counted and numbered; each account behavior corresponds to a unique and definite number; Obtain the user's behavior information in real time according to the permissions granted by the user, and query the corresponding number; At every preset time, the behavior information of all users within the time is counted, the location information of the users is obtained, and a behavior matrix is ​​constructed according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information; Arrange the behavior matrix in chronological order, extract all matrix data of the same account label based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve; Based on the behavior curve, the accounts are clustered by time period to obtain similar accounts in each time period; The step of counting the behavior information of all users within a preset time period, obtaining the user's location information, and constructing a behavior matrix according to the location information includes: Obtain the number of access accounts in real time, arrange the number of accounts according to the time axis, and fit the number change function; Perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function; Intercept the quantity change function within a period, segment the quantity change function within a period, and determine the statistical nodes; If the current moment reaches the statistical node, the behavior information of all users in the previous time period is counted, the user's location information is obtained, and a behavior matrix is ​​constructed according to the location information.

2. The account management method based on big data according to claim 1, characterized in that: If the current moment reaches the statistical node, the steps of statistically analyzing the behavior information of all users in the previous time period, obtaining the user's location information, and constructing a behavior matrix according to the location information include: Obtain the previous statistical node of the current statistical node, and use the time period from the previous statistical node to the current statistical node as the time period to be counted; For any account within the statistical time period, obtain behavior information in real time and query the number corresponding to the behavior information; Insert the account tag inside the number as the element value; Acquire the location information of the account, determine the row and column position according to the location information, and create an element according to the row and column position and the element value; Count the elements of all accounts and get the behavior matrix.

3. The account management method based on big data according to claim 1, characterized in that: The steps of arranging the behavior matrix in chronological order, extracting all matrix data of the same account label based on the values ​​in the row and column positions, fitting the matrix data, and creating a behavior curve include: Arrange the behavior matrix in chronological order; Selecting row and column positions in the bottom-level behavior matrix in turn, and querying the account labels in the values ​​at the row and column positions; the bottom-level behavior matrix is ​​the behavior matrix at the beginning of the duration; Locate elements in other behavior matrices based on account labels, and use the located element positions and element values ​​as matrix data; Fit the matrix data to obtain the behavior curve corresponding to the account label.

4. The account management method based on big data according to claim 3, characterized in that: The step of clustering the accounts by time period based on the behavior curve to obtain similar accounts in each time period comprises: Pair account tags in pairs, and read the behavior curve of any two paired accounts based on the account tags; Place the behavior curves on the same horizontal axis and calculate the curve distance; All accounts are clustered according to the curve distance to obtain similar accounts within the current time period.

5. The account management method based on big data according to claim 4, characterized in that: The generation process of the behavior curve is: Randomly select four matrix data and draw Bezier curves based on the four matrix data; wherein the number of random selections is a preset value; Calculate the mean distance between any Bezier curve and other Bezier curves, and select the Bezier curve corresponding to the minimum distance mean as the final behavior curve; The Bezier curve is: B(α)=(1-α) 3 P0+3(1―α) 2 αP1+3(1―α)α 2 P2+α 3 P3; where P0, P1, P2 and P3 are four selected matrix data, α is a preset parameter, α∈[0,1]; The calculation process of the curve distance is: In the formula, f(x) is the function expression of one of the curves based on the same coordinate axis, g(x) is the function expression of the other curve based on the same coordinate axis; x0 is the common left endpoint of the two curves, and x1 is the right endpoint of the curve with a shorter horizontal axis distance on the same coordinate axis; Place the two curves on the same coordinate axis and align the left endpoints, then calculate the integral of the curve difference. The part of the integral to be calculated needs to be the relatively overlapping part of the two curves on the same coordinate axis. Therefore, x0 is the common left endpoint of the two curves, and x1 is the right endpoint of the curve with a shorter horizontal axis distance on the same coordinate axis, which is used to ensure that both curves have values ​​during the integral calculation process.

6. An account management system based on big data, characterized in that: The system comprises: The behavior numbering module is used to count all account behaviors in the preprocessing stage and number the account behaviors; each account behavior corresponds to a unique and definite number; The number query module is used to obtain the user's behavior information in real time according to the permissions granted by the user and query the corresponding number; A behavior matrix construction module is used to count the behavior information of all users within a preset time period, obtain the user's location information, and construct a behavior matrix according to the location information; the row and column positions of the behavior matrix correspond to the location information, and the value of each row and column position is determined by the account tag and the number of the behavior information; The curve fitting module is used to arrange the behavior matrix in chronological order, extract all matrix data of the same account label based on the values ​​in the row and column positions, fit the matrix data, and create a behavior curve; An account clustering module, used to cluster accounts by time period based on the behavior curve to obtain similar accounts in each time period; The behavior matrix building blocks include: Function fitting unit, used to obtain the number of access accounts in real time, arrange the number of accounts according to the time axis, and fit the number change function; A period extraction unit is used to perform Fourier transform on the quantity change function, analyze the amplitude of each frequency component, and select the period corresponding to the maximum amplitude as the period of the quantity change function; A function segmentation unit is used to intercept the quantity change function within a period, segment the quantity change function within a period, and determine the statistical nodes; An execution unit is constructed to count the behavior information of all users in the previous time period if the current moment reaches the statistical node, obtain the user's location information, and construct a behavior matrix according to the location information.

7. The big data-based account management system according to claim 6, characterized in that: The curve fitting module comprises: A matrix arrangement unit, used to arrange the behavior matrix according to the time sequence; A label query unit is used to select row and column positions in the bottom-level behavior matrix in sequence, and query the account labels in the values ​​at the row and column positions; the bottom-level behavior matrix is ​​the behavior matrix at the beginning of the duration; A data locating unit is used to locate elements in other behavior matrices in sequence based on account labels, and use the located element positions and element values ​​as matrix data; The fitting execution unit is used to fit the matrix data to obtain the behavior curve corresponding to the account label.

8. The big data-based account management system according to claim 6, characterized in that: The account clustering module includes: An account pairing unit is used to pair account tags in pairs and read the behavior curve of any two paired accounts according to the account tags; A distance calculation unit, used to place the behavior curves in the same horizontal axis and calculate the curve distance; The clustering execution unit is used to cluster all accounts according to the curve distance to obtain similar accounts within the current time period.

Images