A billing method and system based on traffic usage

Abstract

The present invention discloses a billing method and system based on traffic usage. According to user attributes, the user traffic usage corresponding to a time point is mapped to a three-dimensional space. This facilitates the subsequent analysis of the correlation between time points, user attributes and traffic. Converting a three-dimensional application traffic curve, an identity traffic curve and a regional traffic curve into a two-dimensional traffic spectrum can facilitate convolution calculations. In the two-dimensional traffic spectrum, the correlation between the user-specific time points and user attributes and the user packages is predicted to obtain an association probability matrix. The first Bayesian network capable of predicting the optimal package for the user is constructed through the association probability matrix. The technical effect of achieving the flow billing that satisfies the user is achieved by judging the probability of the traffic used by each user attribute of the user's traffic usage, and allocating it to a user package that can be used universally but is less used in the subsequent user packages.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a flow usage-based billing method and system. Background Art

[0002] At present, when using traffic, the corresponding traffic packages are different according to the identity, time, region and application of the traffic. There may be a situation where one traffic package has been consumed and another traffic package has not been consumed. If you continue to consume traffic under the identity, time, region and application conditions corresponding to the traffic package that has been consumed, you will be charged separately. Therefore, it is necessary to classify the packages corresponding to the traffic used by users and find the most suitable traffic package to meet the needs of using the most traffic with the least consumption. Summary of the invention

[0003] The purpose of the present invention is to provide a method and system for charging based on traffic usage to solve the above problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a method for charging based on traffic usage, including:

[0005] Extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes;

[0006] Based on the user traffic and user package, mapping is performed into a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map, and a region three-dimensional point map;

[0007] Based on the application three-dimensional point map, identity three-dimensional point map and region three-dimensional point map, the flow change state is judged to obtain the application flow curve map, identity flow curve map and region flow curve map;

[0008] Based on the application traffic curve graph, the identity traffic curve graph and the regional traffic curve graph, a two-dimensional traffic spectrum graph is obtained;

[0009] Based on the two-dimensional traffic spectrum, features are extracted through a probability discriminant network to obtain a correlation probability matrix;

[0010] Based on the association probability matrix, construct a first Bayesian network;

[0011] Through the first Bayesian network, it is determined which user package the user's traffic usage belongs to at multiple time points, and the traffic fee is obtained.

[0012] Optionally, constructing a first Bayesian network based on the association probability matrix includes:

[0013] The user package is regarded as a child node of time point, application attribute, identity attribute and region attribute;

[0014] Each row of the association probability matrix is ​​added to form a time association probability matrix; the time association probability matrix is ​​used as the association relationship between the time point and the user package;

[0015] Add each column corresponding to the application attribute to obtain the application attribute association probability matrix; use the application attribute association probability matrix as the association relationship between the application attribute and the user package;

[0016] Add each column corresponding to the identity attribute to obtain the identity attribute association probability matrix; use the identity attribute association probability matrix as the association relationship between the identity attribute and the user package;

[0017] Each column corresponding to the regional attribute is added to obtain the regional attribute association probability matrix; the regional attribute association probability matrix is ​​used as the association relationship between the regional attribute and the user package.

[0018] Optionally, based on the two-dimensional traffic spectrum, extracting features through a probability discriminant network to obtain an associated probability matrix includes:

[0019] Convolve the two-dimensional convolution kernel in the width and length directions of the two-dimensional spectrum with a step size of 1, detect the correlation characteristics of the flow in some areas, and obtain the regional correlation features;

[0020] Convolve the first one-dimensional convolution kernel in the long direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the flow in time and obtain the time correlation characteristics;

[0021] Convolve the second one-dimensional convolution kernel in the width direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the traffic on the user attributes and obtain the attribute correlation characteristics;

[0022] Based on the region association features, time association features and attribute association features, an association probability matrix is ​​obtained through a probability discriminant network.

[0023] Optionally, obtaining a two-dimensional traffic spectrum based on the application traffic curve graph, the identity traffic curve graph and the regional traffic curve graph includes:

[0024] The application flow curve graph, the identity flow curve graph and the regional flow curve graph are sequentially superimposed in the direction of user attributes to obtain a three-dimensional user graph;

[0025] Mapping the height of the value that is 1 in the three-dimensional user graph to the initial two-dimensional image to obtain a two-dimensional flow spectrum;

[0026] The two-dimensional traffic spectrum is a grayscale image. The larger the grayscale value in the two-dimensional traffic spectrum, the more traffic the user uses. The length of the two-dimensional traffic spectrum represents a time point, and the width represents a user attribute.

[0027] Optionally, the height of the first one-dimensional convolution kernel is equal to the width of the two-dimensional flow spectrum, and the width is equal to 1; the height of the second one-dimensional convolution kernel is equal to 1, and the width is equal to the length of the two-dimensional flow spectrum.

[0028] Optionally, the obtaining of an association probability matrix based on the region association features, time association features and attribute association features through a probability discriminant network includes:

[0029] The probability discrimination network includes a first time neural network, a first attribute neural network and a first regional convolutional network;

[0030] Inputting the time correlation feature into a first time neural network, detecting the correlation probability between multiple time points and traffic, and obtaining a time correlation probability vector;

[0031] Inputting the attribute association feature into a first attribute neural network, detecting the association probability between multiple user attributes and traffic, and obtaining a user attribute association probability vector;

[0032] Multiplying the time association probability vector by the transpose of the user attribute association probability vector to obtain a first probability matrix;

[0033] Inputting the region association features into the first regional convolutional network to obtain a second probability matrix;

[0034] The values ​​of corresponding positions in the first probability matrix and the second probability matrix are averaged to obtain a relationship probability matrix.

[0035] Optionally, mapping the user traffic and user packages to a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map, and a region three-dimensional point map includes:

[0036] Evenly divide the user's traffic usage to obtain multiple traffic segments;

[0037] A three-dimensional point graph is constructed with multiple time points as horizontal coordinates, a user attribute as vertical coordinates, and multiple traffic segments as vertical coordinates;

[0038] Among them, if the user attribute is an application attribute, a corresponding application 3D point map is obtained; if the user attribute is an identity attribute, a corresponding identity 3D point map is obtained; if the user attribute is a region attribute, a corresponding region 3D point map is obtained;

[0039] Fill the user usage traffic at multiple time points into the application three-dimensional dot map, identity three-dimensional dot map and region three-dimensional dot map respectively.

[0040] Optionally, the determining, by the first Bayesian network, to which user packages the user's traffic usage belongs at multiple time points to obtain traffic charges includes:

[0041] According to the user package, the total package traffic volume is obtained;

[0042] According to the Bayesian network, determine which type of package the user uses and obtain the package traffic used;

[0043] Compare the package traffic used with the total package traffic to determine whether it exceeds the total package traffic;

[0044] If it exceeds the limit, the excess traffic will be charged. If it does not exceed the limit, the total traffic of the package will be charged.

[0045] Optionally, judging the flow change state based on the application three-dimensional point map, the identity three-dimensional point map and the region three-dimensional point map to obtain the application flow curve map, the identity flow curve map and the region flow curve map includes:

[0046] Fitting a plurality of points in the application three-dimensional point graph into an application flow curve to obtain an application flow curve graph;

[0047] Fitting multiple points in the identity three-dimensional point graph into an identity flow curve to obtain an identity flow curve graph;

[0048] A plurality of points in the regional three-dimensional point map are fitted into a regional flow curve to obtain a regional flow curve map.

[0049] In a second aspect, an embodiment of the present invention provides a billing system based on traffic usage, including:

[0050] Acquisition module: extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes;

[0051] Mapping module: based on the user's traffic usage and user package, mapping to the three-dimensional space to obtain the application three-dimensional point map, identity three-dimensional point map and region three-dimensional point map;

[0052] Curve construction module: based on the application three-dimensional point map, identity three-dimensional point map and regional three-dimensional point map, determine the flow change state, and obtain the application flow curve map, identity flow curve map and regional flow curve map;

[0053] Spectrum module: based on the application flow curve, identity flow curve and regional flow curve, obtain a two-dimensional flow spectrum;

[0054] Relationship probability module: Based on the two-dimensional traffic spectrum, the probability discriminant network is used to extract features and obtain a correlation probability matrix;

[0055] Bayesian network module: constructing a first Bayesian network based on the association probability matrix;

[0056] Traffic detection module: through the first Bayesian network, determine the user package to which the user's traffic usage belongs at multiple time points and obtain the traffic fee.

[0057] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0058] The embodiment of the present invention also provides a billing method and system based on traffic usage, the method comprising: extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages comprising exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes comprising application attributes, identity attributes and region attributes; based on the user traffic usage and user packages, mapping to three-dimensional space to obtain application three-dimensional point graphs, identity three-dimensional point graphs and region three-dimensional point graphs; based on the application three-dimensional point graphs, identity three-dimensional point graphs and region three-dimensional point graphs, judging the traffic change state to obtain application traffic curve graphs, identity traffic curve graphs and region traffic curve graphs; based on the application traffic curve graphs, identity traffic curve graphs and region traffic curve graphs, obtaining a two-dimensional traffic spectrum; based on the two-dimensional traffic spectrum, extracting features through a probability discriminant network to obtain an associated probability matrix; based on the associated probability matrix, constructing a first Bayesian network; through the first Bayesian network, judging the user packages to which the user traffic usage belongs at multiple time points to obtain traffic charges.

[0059] In the present invention, the user usage flow corresponding to the time point is mapped to a three-dimensional space according to the user attributes. This facilitates the subsequent analysis of the correlation between the time point, user attributes and flow. The three-dimensional application point map, the three-dimensional identity point map and the three-dimensional regional point map are respectively converted into an application flow curve map, an identity flow curve map and a regional flow curve map. The amount of data can be increased, which is convenient for the subsequent detection and calculation of big data. The three-dimensional application flow curve map, the identity flow curve map and the regional flow curve map are converted into a two-dimensional flow spectrum map, which can facilitate convolution calculation. In the two-dimensional flow spectrum map, the correlation between the user-specific time point and user attributes and the user package is predicted to obtain an association probability matrix. The first Bayesian network that can predict the optimal package for the user is constructed through the association probability matrix. The technical effect of judging the probability of the flow used by each user attribute of the user's flow and allocating it to the user package that can be used universally but less used in the subsequent user package is achieved, so that the flow billing satisfies the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The present invention provides a method for charging based on flow usage.

[0061] Figure 2 It is a structural diagram of a first Bayesian network in a traffic usage-based billing method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0063] Example 1

[0064] like Figure 1 As shown, an embodiment of the present invention provides a billing method based on traffic usage, the method comprising:

[0065] S101: Extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes.

[0066] The user usage flow is the flow corresponding to the user attributes.

[0067] Among them, the identity indicates whether the person is a student, which is used to determine whether the corresponding identity-exclusive package, that is, the campus-exclusive package, can be used.

[0068] The application attribute includes multiple applications using traffic, the identity attribute includes two attributes: identity is a student and identity is not a student, and the region attribute includes multiple regions corresponding to the traffic operator using traffic. In this embodiment, the regions are divided into multiple regions based on provinces.

[0069] S102: Mapping the user traffic usage and user package to a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map, and a region three-dimensional point map;

[0070] S103: Based on the application three-dimensional point map, the identity three-dimensional point map and the region three-dimensional point map, determine the flow change state, and obtain an application flow curve map, an identity flow curve map and a region flow curve map;

[0071] S104: Based on the region association features, time association features and attribute association features, an association probability matrix is ​​obtained through a probability discriminant network.

[0072] The association probability matrix includes a matrix of associations corresponding to the user's traffic usage. Multiple applications in the application attribute, multiple identities in the identity attribute, and region attributes

[0073] S105: Constructing a first Bayesian network based on the association probability matrix.

[0074] The Bayesian network is used to classify the traffic used by users and determine which user package's traffic is used at multiple time points.

[0075] S106: Using the first Bayesian network, determine the user packages to which the user's traffic usage belongs at multiple time points, and obtain traffic charges.

[0076] The traffic fee refers to the total fee for the user to use the traffic.

[0077] The first Bayesian network is trained using a universal but less commonly used user package. For example, if a user uses a video application from 23:00 to 7:00, he can use either a dedicated time package or a dedicated application package. However, because the traffic used by users at night is less than that used by video applications 24 hours a day, the dedicated time package is used as the labeled data for training.

[0078] Optionally, constructing a first Bayesian network based on the association probability matrix includes:

[0079] The user package is treated as a child node of the time point, application attribute, identity attribute, and region attribute.

[0080] Among them, the first Bayesian network can detect the user packages corresponding to the traffic usage of multiple users at different time points.

[0081] Wherein, the structure of the first Bayesian network is as follows: Figure 2 shown.

[0082] Each row of the association probability matrix is ​​added to form a time association probability matrix; the time association probability matrix is ​​used as the association relationship between the time point and the user package;

[0083] Add each column corresponding to the application attribute to obtain the application attribute association probability matrix; use the application attribute association probability matrix as the association relationship between the application attribute and the user package;

[0084] Add each column corresponding to the identity attribute to obtain the identity attribute association probability matrix; use the identity attribute association probability matrix as the association relationship between the identity attribute and the user package;

[0085] Each column corresponding to the regional attribute is added to obtain the regional attribute association probability matrix; the regional attribute association probability matrix is ​​used as the association relationship between the regional attribute and the user package.

[0086] Optionally, based on the two-dimensional traffic spectrum, extracting features through a probability discriminant network to obtain an associated probability matrix includes:

[0087] Convolve the two-dimensional convolution kernel in the width and length directions of the two-dimensional spectrum with a step size of 1, detect the correlation characteristics of the flow in some areas, and obtain the regional correlation features;

[0088] Convolve the first one-dimensional convolution kernel in the long direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the flow in time and obtain the time correlation characteristics;

[0089] Convolve the second one-dimensional convolution kernel in the width direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the traffic on the user attributes and obtain the attribute correlation characteristics;

[0090] Based on the region association features, time association features and attribute association features, an association probability matrix is ​​obtained through a probability discriminant network.

[0091] Optionally, obtaining a two-dimensional traffic spectrum based on the application traffic curve graph, the identity traffic curve graph and the regional traffic curve graph includes:

[0092] The application flow curve graph, the identity flow curve graph and the regional flow curve graph are sequentially superimposed in the direction of user attributes to obtain a three-dimensional user graph.

[0093] The length of the three-dimensional user graph is used as the horizontal coordinate, the width is used as the vertical coordinate, and the height is used as the vertical coordinate. The horizontal coordinate of the three-dimensional user graph represents the time point, the vertical coordinate represents the user's usage flow, and the vertical coordinate represents the user attribute. The points on the vertical coordinate of the three-dimensional user graph represent application attributes, identity attributes, and regional attributes respectively.

[0094] Obtain an initial two-dimensional image having a length equal to the length of the three-dimensional user graph and a width equal to the width of the three-dimensional user graph;

[0095] The height of the value of 1 in the three-dimensional user map is mapped in the initial two-dimensional image to obtain a two-dimensional flow spectrum.

[0096] The two-dimensional traffic spectrum is equivalent to compressing the three-dimensional user image, because the traffic usage of a user with an identity attribute using an application in a region from one time point to the previous time point is fixed and unique, so it can be compressed.

[0097] The height of the value of 1 in the three-dimensional user graph is taken as the flow value, the highest point in the vertical coordinate of the three-dimensional user graph is taken as the total flow value, and the flow value is divided by the total flow value and multiplied by 255 to obtain the gray value.

[0098] The two-dimensional traffic spectrum is a grayscale image. The larger the grayscale value in the two-dimensional traffic spectrum, the more traffic the user uses. The length of the two-dimensional traffic spectrum represents a time point, and the width represents a user attribute.

[0099] Optionally, the height of the first one-dimensional convolution kernel is equal to the width of the two-dimensional flow spectrum, and the width is equal to 1; the height of the second one-dimensional convolution kernel is equal to 1, and the width is equal to the length of the two-dimensional flow spectrum.

[0100] The convolution kernel includes height, width and number of channels. For example, 3×5 convolution means that the height of the convolution kernel is 3 and the width is 5. The length of the two-dimensional spectrum and the width of the convolution kernel are in the same direction during convolution, and the width of the two-dimensional spectrum and the width of the convolution kernel are in the same direction during convolution.

[0101] Optionally, the obtaining of an association probability matrix based on the region association features, time association features and attribute association features through a probability discriminant network includes:

[0102] Optionally, the obtaining of an association probability matrix based on the region association features, time association features and attribute association features through a probability discriminant network includes:

[0103] The probability discrimination network includes a first time neural network, a first attribute neural network and a first regional convolutional network.

[0104] The time correlation feature is input into a first time neural network, the correlation probability between multiple time points and traffic is detected, and a time correlation probability vector is obtained.

[0105] The time association probability vector represents the proportion of the traffic used by the user from one time point to the previous time point of each day to the traffic used by the user. During training, the average value of the traffic used from one time point to the previous time point in the history of the user's traffic used is divided by the quotient of the user's traffic used as the annotation data.

[0106] Among them, the first time neural network is a fully connected neural network (Fully Connected Neural Network, FCN), user.

[0107] The attribute association feature is input into a first attribute neural network, the association probability between multiple user attributes and traffic is detected, and a user attribute association probability vector is obtained.

[0108] The user attribute association probability vector represents the proportion of traffic used by multiple applications to the user's traffic, the proportion of traffic used by different identities to the user's traffic, and the proportion of traffic used by different regions to the user's traffic.

[0109] Among them, the first attribute neural network is a fully connected neural network (Fully Connected Neural Network, FCN).

[0110] The time-related probability vector is multiplied by the transpose of the user attribute-related probability vector to obtain a first probability matrix, wherein the first probability matrix represents the relationship between the traffic of different applications used by users with different identity attributes in different regions.

[0111] The region association features are input into the first regional convolutional network to obtain a second probability matrix.

[0112] Among them, the first regional convolutional network is a convolutional neural network (Convolutional Neural Networks, CNN).

[0113] The second probability matrix is ​​a relationship between the traffic of different applications used by users with different identity attributes in different regions, which is predicted directly by the first regional convolutional network.

[0114] The values ​​of corresponding positions in the first probability matrix and the second probability matrix are averaged to obtain a relationship probability matrix.

[0115] The number of rows of the first probability matrix is ​​equal to the number of rows of the second probability matrix, and the number of columns of the first probability matrix is ​​equal to the number of columns of the second probability matrix. The value corresponding to the first row and first column of the first probability matrix is ​​averaged with the value corresponding to the first row and first column of the second probability matrix to obtain a relationship probability matrix.

[0116] Optionally, mapping the user traffic and user packages to a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map, and a region three-dimensional point map includes:

[0117] The user's usage traffic is evenly divided to obtain multiple traffic segments.

[0118] The traffic segmentation means dividing the user's traffic into multiple segments. For example, if the user uses 5G of traffic, the user's traffic is divided into 5 parts equally, with [0-1)G as the first traffic segment, [1-2)G as the second traffic segment, [2-3)G as the third traffic segment, [3-4)G as the fourth traffic segment, and [4-5]G as the fifth traffic segment.

[0119] A three-dimensional point graph is constructed with multiple time points as the horizontal coordinates, a user attribute as the vertical coordinate, and multiple traffic segments as the vertical coordinates.

[0120] Among them, if the user attribute is an application attribute, a corresponding application three-dimensional point map is obtained; if the user attribute is an identity attribute, a corresponding identity three-dimensional point map is obtained; if the user attribute is a regional attribute, a corresponding regional three-dimensional point map is obtained.

[0121] Among them, the application three-dimensional point map is constructed with multiple time points as the horizontal coordinates, the application attributes as the vertical coordinates, and the multiple traffic segments as the vertical coordinates. The identity three-dimensional point map is constructed with multiple time points as the horizontal coordinates, the identity attributes as the vertical coordinates, and the multiple traffic segments as the vertical coordinates. The regional three-dimensional point map is constructed with multiple time points as the horizontal coordinates, the regional attributes as the vertical coordinates, and the multiple traffic segments as the vertical coordinates.

[0122] Fill the user usage traffic at multiple time points into the application three-dimensional dot map, identity three-dimensional dot map and region three-dimensional dot map respectively.

[0123] For example, according to the time period from 14:00 to 15:00, the user traffic used by the video application in region A by the user with the identity of a student is 0.3G, and the position of the horizontal coordinate corresponding to the time point 15:00 in the application three-dimensional dot graph, the vertical coordinate corresponding to the video application, and the vertical coordinate corresponding to the first traffic segment are set to 1. The horizontal coordinate corresponding to the time point 15:00 in the identity three-dimensional dot graph, the vertical coordinate corresponding to the identity of the student, and the vertical coordinate corresponding to the first traffic segment are set to 1. The horizontal coordinate corresponding to the time point 15:00 in the region three-dimensional dot graph, the vertical coordinate corresponding to region A, and the vertical coordinate corresponding to the first traffic segment are set to 1.

[0124] Optionally, the determining, by the first Bayesian network, to which user packages the user's traffic usage belongs at multiple time points to obtain traffic charges includes:

[0125] According to the user package, the total package traffic volume is obtained.

[0126] The total package traffic volume refers to the traffic volume of the package purchased by the user, such as the exclusive time package 5G that can only be used from 11 pm to 6 am, the exclusive application package 5G that can only be used for video applications, the identity exclusive package 5G that can only be used by students, and the regional restriction package 5G that can only be used locally.

[0127] According to the Bayesian network, it is determined which type of package the user uses, and the package traffic used is obtained.

[0128] Among them, there are 3G exclusive time packages, 2G exclusive application packages that can only be used for video applications, 5G identity-exclusive packages that can only be used by students, and 5G region-restricted packages that can only be used locally.

[0129] Compare the package traffic used with the total package traffic to determine whether it exceeds the total package traffic;

[0130] If it exceeds the limit, the excess traffic will be charged. If it does not exceed the limit, the total traffic of the package will be charged.

[0131] Optionally, judging the flow change state based on the application three-dimensional point map, the identity three-dimensional point map and the region three-dimensional point map to obtain the application flow curve map, the identity flow curve map and the region flow curve map includes:

[0132] A plurality of points in the application three-dimensional point graph are fitted into an application flow curve to obtain an application flow curve graph.

[0133] Among them, polynomial fitting is used for fitting.

[0134] A plurality of points in the identity three-dimensional point graph are fitted into an identity flow curve to obtain an identity flow curve graph.

[0135] A plurality of points in the regional three-dimensional point map are fitted into a regional flow curve to obtain a regional flow curve map.

[0136] The application flow curve graph, identity flow curve graph and regional flow curve graph are all three-dimensional images.

[0137] Example 2

[0138] Based on the above-mentioned billing method based on traffic usage, an embodiment of the present invention further provides a billing system based on traffic usage, the system comprising:

[0139] The acquisition module is used to extract user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes.

[0140] The mapping module is used to map the user's traffic usage and user package to a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map and a regional three-dimensional point map.

[0141] The curve construction module is used to determine the flow change state based on the application three-dimensional point map, identity three-dimensional point map and regional three-dimensional point map, and obtain the application flow curve map, identity flow curve map and regional flow curve map.

[0142] The spectrum module is used to obtain a two-dimensional traffic spectrum based on the application traffic curve, identity traffic curve and regional traffic curve.

[0143] The relationship probability module is used to extract features based on the two-dimensional traffic spectrum through a probability discriminant network to obtain a correlation probability matrix.

[0144] The Bayesian network module is used to construct a first Bayesian network based on the association probability matrix.

[0145] The traffic detection module is used to determine the user package to which the user's traffic usage belongs at multiple time points through the first Bayesian network, and obtain the traffic fee.

Claims

1. A billing method based on traffic usage, characterized in that: include: Extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes; Based on the user traffic and user package, mapping is performed into a three-dimensional space to obtain an application three-dimensional point map, an identity three-dimensional point map, and a region three-dimensional point map; Based on the application three-dimensional point map, identity three-dimensional point map and region three-dimensional point map, the flow change state is judged to obtain the application flow curve map, identity flow curve map and region flow curve map; Based on the application traffic curve graph, the identity traffic curve graph and the regional traffic curve graph, a two-dimensional traffic spectrum graph is obtained; Based on the two-dimensional traffic spectrum, features are extracted through a probability discriminant network to obtain a correlation probability matrix; Based on the association probability matrix, construct a first Bayesian network; Through the first Bayesian network, it is determined which user package the user's traffic usage belongs to at multiple time points, and the traffic fee is obtained.

2. A method for charging based on traffic usage according to claim 1, characterized in that: The step of constructing a first Bayesian network based on the association probability matrix includes: The user package is regarded as a child node of time point, application attribute, identity attribute and region attribute; Each row of the association probability matrix is ​​added to form a time association probability matrix; the time association probability matrix is ​​used as the association relationship between the time point and the user package; Add each column corresponding to the application attribute to obtain the application attribute association probability matrix; use the application attribute association probability matrix as the association relationship between the application attribute and the user package; Add each column corresponding to the identity attribute to obtain the identity attribute association probability matrix; use the identity attribute association probability matrix as the association relationship between the identity attribute and the user package; Each column corresponding to the regional attribute is added to obtain the regional attribute association probability matrix; the regional attribute association probability matrix is ​​used as the association relationship between the regional attribute and the user package.

3. A method for charging based on traffic usage according to claim 1, characterized in that: The method of extracting features based on the two-dimensional traffic spectrum through a probability discriminant network to obtain an associated probability matrix includes: Convolve the two-dimensional convolution kernel in the width and length directions of the two-dimensional spectrum with a step size of 1, detect the correlation characteristics of the flow in some areas, and obtain the regional correlation features; Convolve the first one-dimensional convolution kernel in the long direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the flow in time and obtain the time correlation characteristics; Convolve the second one-dimensional convolution kernel in the width direction of the two-dimensional spectrum with a step size of 1 to detect the correlation characteristics of the traffic on the user attributes and obtain the attribute correlation characteristics; Based on the region association features, time association features and attribute association features, an association probability matrix is ​​obtained through a probability discriminant network.

4. A method for charging based on traffic usage according to claim 1, characterized in that: The two-dimensional flow spectrum is obtained based on the application flow curve, the identity flow curve and the regional flow curve, including: The application flow curve graph, the identity flow curve graph and the regional flow curve graph are sequentially superimposed in the direction of user attributes to obtain a three-dimensional user graph; Mapping the height of the value that is 1 in the three-dimensional user graph to the initial two-dimensional image to obtain a two-dimensional flow spectrum; The two-dimensional traffic spectrum is a grayscale image. The larger the grayscale value in the two-dimensional traffic spectrum, the more traffic the user uses. The length of the two-dimensional traffic spectrum represents a time point, and the width represents a user attribute.

5. A method for charging based on traffic usage according to claim 3, characterized in that: The height of the first one-dimensional convolution kernel is equal to the width of the two-dimensional flow spectrum, and the width is equal to 1; the height of the second one-dimensional convolution kernel is equal to 1, and the width is equal to the length of the two-dimensional flow spectrum.

6. A method for charging based on traffic usage according to claim 3, characterized in that: The method of obtaining an association probability matrix based on the regional association features, time association features and attribute association features through a probability discriminant network includes: The probability discrimination network includes a first time neural network, a first attribute neural network and a first regional convolutional network; Inputting the time correlation feature into a first time neural network, detecting the correlation probability between multiple time points and traffic, and obtaining a time correlation probability vector; Inputting the attribute association feature into a first attribute neural network, detecting the association probability between multiple user attributes and traffic, and obtaining a user attribute association probability vector; Multiplying the time association probability vector by the transpose of the user attribute association probability vector to obtain a first probability matrix; Inputting the region association features into the first regional convolutional network to obtain a second probability matrix; The values ​​of corresponding positions in the first probability matrix and the second probability matrix are averaged to obtain a relationship probability matrix.

7. A method for charging based on traffic usage according to claim 1, characterized in that: The mapping to the three-dimensional space based on the user traffic and user package to obtain the application three-dimensional point map, identity three-dimensional point map and region three-dimensional point map includes: Evenly divide the user's traffic usage to obtain multiple traffic segments; A three-dimensional point graph is constructed with multiple time points as horizontal coordinates, a user attribute as vertical coordinates, and multiple traffic segments as vertical coordinates; Among them, if the user attribute is an application attribute, a corresponding application 3D point map is obtained; if the user attribute is an identity attribute, a corresponding identity 3D point map is obtained; if the user attribute is a region attribute, a corresponding region 3D point map is obtained; Fill the user usage traffic at multiple time points into the application three-dimensional dot map, identity three-dimensional dot map and region three-dimensional dot map respectively.

8. A method for charging based on traffic usage according to claim 1, characterized in that: The first Bayesian network is used to determine the user packages to which the user's traffic usage belongs at multiple time points, and the traffic fee is obtained, including: According to the user package, the total package traffic volume is obtained; According to the Bayesian network, determine which type of package the user uses and obtain the package traffic used; Compare the package traffic used with the total package traffic to determine whether it exceeds the total package traffic; If it exceeds the limit, the excess traffic will be charged. If it does not exceed the limit, the total traffic of the package will be charged.

9. A method for charging based on traffic usage according to claim 1, characterized in that: The method of judging the flow change state based on the application three-dimensional point map, the identity three-dimensional point map and the region three-dimensional point map to obtain the application flow curve map, the identity flow curve map and the region flow curve map includes: Fitting a plurality of points in the application three-dimensional point graph into an application flow curve to obtain an application flow curve graph; Fitting multiple points in the identity three-dimensional point graph into an identity flow curve to obtain an identity flow curve graph; A plurality of points in the regional three-dimensional point map are fitted into a regional flow curve to obtain a regional flow curve map.

10. A billing system based on traffic usage, characterized in that: include: Acquisition module: extracting user attributes, user traffic usage corresponding to multiple time points and user packages; the user packages include exclusive time packages, exclusive application packages, identity exclusive packages and region restricted packages; the user attributes include application attributes, identity attributes and region attributes; Mapping module: based on the user's traffic usage and user package, mapping to three-dimensional space to obtain application three-dimensional point map, identity three-dimensional point map and region three-dimensional point map; Curve construction module: based on the application three-dimensional point map, identity three-dimensional point map and regional three-dimensional point map, determine the flow change state, and obtain the application flow curve map, identity flow curve map and regional flow curve map; Spectrum module: based on the application flow curve, identity flow curve and regional flow curve, obtain a two-dimensional flow spectrum; Relationship probability module: Based on the two-dimensional traffic spectrum, the probability discriminant network is used to extract features and obtain a correlation probability matrix; Bayesian network module: constructing a first Bayesian network based on the association probability matrix; Traffic detection module: through the first Bayesian network, determine the user package to which the user's traffic usage belongs at multiple time points and obtain the traffic fee.

Images