Traffic user data analysis method and system

Abstract

The invention discloses a traffic user data analysis method and system. And carrying out segmentation according to the time point when the user stops using the traffic, and only retaining the time point when the traffic is used and the corresponding used traffic to obtain a traffic curve graph and a traffic discrete graph. The flow curve graph represents that a plurality of pieces of continuously used flow data are fitted into a curve. And according to the horizontal coordinate corresponding to the fitted curve, cutting the three-dimensional flow curve graph to obtain a plurality of time discrete graphs. And inputting the multiple time discrete diagrams into a time convolution network, and carrying out convolution on the multiple time discrete diagrams and the traffic discrete diagram through a three-dimensional convolution kernel to find a change rule after the traffic is used each time. And by fusing the time flow three-dimensional diagram, the flow curve characteristic diagram and the flow discrete characteristic diagram, the rule of using the flow every day by the user can be found, so that the technical effect of more accurately analyzing the flow used by the user is achieved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a flow user data analysis method and system. Background Art

[0002] At present, with the development of the Internet, all walks of life have also carried out Internet business, and the current use of user traffic data analysis can further analyze the users who use traffic. The current basic implementation method is to analyze and count a type of statistical indicators that users are interested in during the traffic analysis process. The time point when the user stops using traffic, the time point when the user uses traffic, and the characteristics corresponding to the change state of traffic when the user uses traffic have an impact on the user's traffic analysis. In addition, there are certain rules for the use of traffic, so the accuracy of traffic data analysis that does not consider the above characteristics is not enough. Summary of the invention

[0003] The purpose of the present invention is to provide a flow user data analysis method and system 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 analyzing traffic user data, including:

[0005] Obtaining a user traffic usage time point and a corresponding user traffic usage; the user traffic usage time point indicates a time point of traffic usage by the user; the user traffic usage indicates traffic usage by the user at the user traffic usage time point;

[0006] Based on the user traffic usage time points and the corresponding user traffic usage, multiple curves are constructed to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; the traffic dispersion graph represents an image of the user traffic usage marked at multiple user traffic usage time points;

[0007] Based on the flow curve diagram, projecting the characteristics of multiple curves to obtain a flow curve characteristic diagram;

[0008] Based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage, a three-dimensional convolution is performed to obtain a traffic fusion characteristic graph;

[0009] The traffic fusion feature graph is input into a user discrimination network to discriminate the user's state and obtain a user discrimination category; the user discrimination category indicates that the user is classified according to the adaptive traffic.

[0010] Optionally, constructing multiple curves based on the user traffic usage time points and the corresponding user traffic usage to obtain a traffic curve graph and a traffic dispersion graph includes:

[0011] The user traffic usage time point is used as the horizontal axis, and the user traffic usage is used as the vertical axis to obtain a two-dimensional traffic graph;

[0012] Based on the two-dimensional traffic graph, detect the location where the user does not use the traffic and obtain the time point when the traffic is not connected;

[0013] Based on the flow unconnected time point, the two-dimensional flow graph is segmented to obtain a plurality of segmented flow graphs;

[0014] Arrange multiple segmented flow graphs in order from early to late to obtain a flow dispersion graph;

[0015] Fit the points in the segmented flow diagram to obtain a fitting curve;

[0016] The fitting curve is plotted as an image to obtain a two-dimensional flow curve graph; a plurality of segmented flow graphs are correspondingly obtained to obtain a plurality of two-dimensional flow curve graphs;

[0017] Arrange the plurality of two-dimensional flow curve graphs in sequence from early to late to obtain a flow curve graph.

[0018] Optionally, projecting the characteristics of multiple curves based on the flow curve graph to obtain a flow curve characteristic graph includes:

[0019] Taking the length as the dividing point, the side of the flow curve graph is divided multiple times to obtain multiple discrete time graphs; the width of the discrete time graph is equal to the width of the flow curve graph; the length of the discrete time graph is equal to the number of channels of the flow curve graph;

[0020] Convolving the discrete moment graph to extract features and obtain a moment curve feature; multiple discrete moment graphs correspond to multiple moment curve features;

[0021] The multiple time curve features are input into the time convolution network in sequence from early to late according to the time points to obtain a flow curve feature graph.

[0022] Optionally, performing three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage to obtain the traffic fusion characteristic graph includes:

[0023] Divide the user traffic usage time points into 24-hour time intervals to obtain multiple user interval time periods and corresponding user interval traffic;

[0024] The user interval time period is used as the horizontal axis and the corresponding user interval flow rate is used as the vertical axis to draw an image to obtain a user interval two-dimensional image; multiple user interval time periods are corresponding to multiple user interval two-dimensional graphs;

[0025] Multiple user interval two-dimensional graphs are superimposed according to time points from morning to night to obtain a three-dimensional time flow graph;

[0026] Inputting the time flow three-dimensional graph into a three-dimensional convolutional network, extracting the three-dimensional features of each part, and obtaining a time flow three-dimensional feature graph;

[0027] Based on the flow dispersion graph, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature graph; the flow dispersion feature graph is a two-dimensional graph;

[0028] The size of the three-dimensional convolution kernel in the three-dimensional convolution network and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map; the size of the three-dimensional convolution kernel corresponding to the flow discrete map and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map;

[0029] The three-dimensional time flow graph, the flow curve characteristic graph and the flow discrete characteristic graph are superimposed to obtain a flow fusion characteristic graph;

[0030] The flow discrete characteristic diagram and the flow curve characteristic diagram are superimposed on the right side of the flow curve characteristic diagram, adjacent to the end of the curve.

[0031] Optionally, based on the flow dispersion map, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature map, including:

[0032] Obtaining a three-dimensional convolution kernel; the width of the three-dimensional convolution kernel is equal to the width of the flow dispersion map;

[0033] The three-dimensional convolution kernel moves sequentially in the length and channel direction of the flow discrete graph with a step size of 1; the three-dimensional convolution kernel moves sequentially from early to late in time.

[0034] Optionally, the detecting, based on the two-dimensional traffic graph, the location where the user does not use traffic, and obtaining the traffic disconnection time point includes:

[0035] Obtaining a two-dimensional convolution kernel; the width of the two-dimensional convolution kernel is equal to the width of the two-dimensional flow map;

[0036] On the two-dimensional flow map, the two-dimensional convolution kernel is convolved from left to right with a step size of 1 to extract features and obtain a first convolution feature vector;

[0037] The first convolution feature vector is input into the first neural network to detect whether there is flow at a time point, and obtain a flow disconnected time point.

[0038] Optionally, the two-dimensional flow graph is segmented based on the flow unconnected time point to obtain a plurality of segmented flow graphs, including:

[0039] Search from left to right in the two-dimensional flow graph, detect the time point when the flow is not connected, and cut it to obtain a first local segmented flow graph and a first segmented flow graph to be detected; the time point corresponding to the first local segmented flow graph is earlier than the time point corresponding to the first segmented flow graph to be detected;

[0040] Search from left to right in the first segmented traffic graph to be detected, and if a time point is detected that is not a time point when the traffic is not connected, cut it to obtain a second segmented traffic graph to be detected; the starting time point of the second segmented traffic graph to be detected is the time point when the user connects to the traffic;

[0041] Search from left to right in the second segmented flow graph to be detected, detect the time point when the flow is not connected, and cut it to obtain the second local segmented flow graph and the third segmented flow graph to be detected;

[0042] Multiple searches are performed to obtain multiple local segmentation flow graphs; the multiple local segmentation flow graphs include a first local segmentation flow graph and a second local segmentation flow graph;

[0043] The local segmentation flow graph is padded with 0 at both ends to obtain a segmentation flow graph.

[0044] Optionally, one segmented flow graph corresponds to one time point.

[0045] Optionally, the user discrimination network is a convolutional neural network.

[0046] In a second aspect, an embodiment of the present invention provides a flow user data analysis system, including:

[0047] Acquisition module: obtains a user traffic usage time point and a corresponding user traffic usage; the user traffic usage time point indicates a time point of traffic usage by the user; the user traffic usage indicates traffic usage by the user at the user traffic usage time point;

[0048] Curve module: Based on the user traffic usage time points and the corresponding user traffic usage, multiple curves are constructed to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; the traffic dispersion graph represents an image of the user traffic usage marked at multiple user traffic usage time points;

[0049] Curve feature module: based on the flow curve diagram, projecting the features of multiple curves to obtain a flow curve feature diagram;

[0050] Fusion module: performing three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage to obtain a traffic fusion characteristic graph;

[0051] Category module: input the traffic fusion feature map into the user discrimination network, discriminate the user's state, and obtain the user discrimination category; the user discrimination category indicates that the user is classified according to the adaptive traffic.

[0052] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0053] The embodiment of the present invention also provides a flow user data analysis method and system, the method comprising: obtaining a user flow usage time point and a corresponding user flow usage; the user flow usage time point represents a time point of flow usage by the user; the user flow usage represents the flow usage by the user at the user flow usage time point; based on the user flow usage time point and the corresponding user flow usage, constructing multiple curves to obtain a flow curve graph and a flow discrete graph; the flow curve graph represents an image of smooth changes in flow at multiple user flow usage time points; the flow discrete graph represents an image of the user flow usage being annotated at multiple user flow usage time points; based on the flow curve graph, projecting features of multiple curves to obtain a flow curve feature graph; based on the flow discrete graph, the flow curve feature graph, the user flow usage time point and the corresponding user flow usage, performing three-dimensional convolution to obtain a flow fusion feature graph; inputting the flow fusion feature graph into a user discrimination network to discriminate the user's state and obtain a user discrimination category; the user discrimination category represents classifying users according to the adapted flow.

[0054] In the present invention, the flow is segmented according to the time point when the user stops using the flow, and only the time point when the flow is used and the corresponding flow used are retained to obtain a flow curve graph and a flow discrete graph. The flow curve graph represents fitting multiple continuous flow data into a curve. According to the horizontal coordinate corresponding to the fitted curve, the three-dimensional flow curve graph is cut to obtain multiple moment discrete graphs. According to the multiple moment discrete graphs, the time convolution network is input, and the flow discrete graph is convolved through the three-dimensional convolution kernel to find the time length from a time point after the flow is reused to the starting time point of reuse, and then find the time point with the same time length as the time point from the starting time point of the flow when the flow was used last time, so as to find the change of the flow corresponding to the two time points, so as to find the change law after each use of the flow. And by fusing the time flow three-dimensional graph, the flow curve characteristic graph and the flow discrete characteristic graph, the law of the user's daily flow usage can be found, so as to achieve a more accurate technical effect of analyzing the flow used by the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of a flow user data analysis method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0057] Example 1

[0058] like Figure 1 As shown, an embodiment of the present invention provides a method for analyzing traffic user data, the method comprising:

[0059] S101: Obtain a user traffic usage time point and a corresponding user traffic usage; the user traffic usage time point indicates a time point of traffic usage by a user; the user traffic usage indicates traffic usage by a user at the user traffic usage time point.

[0060] S102 constructs multiple curves based on the user traffic usage time points and the corresponding user traffic usage to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; the traffic dispersion graph represents an image of the user traffic usage marked at multiple user traffic usage time points.

[0061] Wherein, the flow curve diagram and the flow dispersion diagram are three-dimensional diagrams.

[0062] S103: Based on the flow curve diagram, the characteristics of multiple curves are projected to obtain a flow curve characteristic diagram.

[0063] Wherein, the flow analysis characteristic graph is a three-dimensional graph.

[0064] S104: Perform three-dimensional convolution based on the flow dispersion graph and the flow curve characteristic graph to obtain a flow fusion characteristic graph.

[0065] Among them, the traffic fusion feature graph is a three-dimensional graph.

[0066] S105: Input the traffic fusion feature graph into a user discrimination network to discriminate the user's state and obtain a user discrimination category; the user discrimination category indicates that the user is classified according to the adaptive traffic.

[0067] Optionally, constructing multiple curves based on the user traffic usage time points and the corresponding user traffic usage to obtain a traffic curve graph and a traffic dispersion graph includes:

[0068] The user traffic usage time point is used as the horizontal axis and the user traffic usage is used as the vertical axis to obtain a two-dimensional traffic graph.

[0069] The two-dimensional traffic graph marks the user traffic usage time points and the points corresponding to the user traffic usage.

[0070] Based on the two-dimensional traffic graph, the location where the user does not use the traffic is detected, and the time point when the traffic is not connected is obtained.

[0071] The position where the user does not use traffic is the user traffic usage time point when the user traffic usage is 0.

[0072] Based on the flow unconnected time point, the two-dimensional flow graph is segmented to obtain a plurality of segmented flow graphs;

[0073] Arrange multiple segmented flow graphs in order from early to late to obtain a flow dispersion graph;

[0074] The points in the segmented flow diagram are fitted to obtain a fitting curve.

[0075] In this embodiment, a polynomial fitting algorithm is used for fitting.

[0076] The fitting curve is plotted as an image to obtain a two-dimensional flow curve diagram; and a plurality of segmented flow diagrams are correspondingly obtained to obtain a plurality of two-dimensional flow curve diagrams.

[0077] Wherein, the fitting curve is drawn using the plot function.

[0078] Wherein, the fitting curve is drawn according to the corresponding abscissa and ordinate in the two-dimensional flow diagram.

[0079] Arrange the plurality of two-dimensional flow curve graphs in sequence from early to late to obtain a flow curve graph.

[0080] Optionally, projecting the characteristics of multiple curves based on the flow curve graph to obtain a flow curve characteristic graph includes:

[0081] Taking the length as the dividing point, the side of the flow curve graph is divided multiple times to obtain multiple discrete time graphs; the width of the discrete time graph is equal to the width of the flow curve graph; the length of the discrete time graph is equal to the number of channels of the flow curve graph.

[0082] The time discrete graph is a two-dimensional graph; the width of the time discrete graph is equal to the width of the flow curve graph.

[0083] Among them, the flow curve chart is obtained by superimposing multiple two-dimensional flow curve charts, so it is a curve when viewed from the front, and the side of the flow curve chart is divided with the length of the flow curve chart as the dividing point, then an image of multiple discrete points viewed from the side is obtained, which is a discrete time chart.

[0084] The time discrete graph represents the flow rates corresponding to multiple curves from a starting point, or the flow rates corresponding to multiple time points with the same time length from the starting point.

[0085] The time discrete graph is convolved to extract features to obtain time curve features; multiple time discrete graphs correspond to multiple time curve features.

[0086] Among them, in this embodiment, convolutional neural networks (CNN) are used for convolution.

[0087] The multiple time curve features are input into the time convolution network in sequence from early to late according to the time points to obtain a flow curve feature graph.

[0088] In this embodiment, a temporal convolutional network (TCN) is used to extract the relationship between traffic changes at multiple time points.

[0089] Optionally, performing three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage to obtain the traffic fusion characteristic graph includes:

[0090] Divide the user traffic usage time points into 24-hour time intervals to obtain multiple user interval time periods and corresponding user interval traffic;

[0091] The user interval time period is used as the horizontal axis and the corresponding user interval flow rate is used as the vertical axis to draw an image to obtain a user interval two-dimensional image; multiple user interval time periods correspond to multiple user interval two-dimensional graphs.

[0092] In this embodiment, the Scatter function can be used to draw discrete points.

[0093] Multiple user interval two-dimensional graphs are superimposed according to time points from morning to night to obtain a three-dimensional time flow graph;

[0094] Inputting the time flow three-dimensional graph into a three-dimensional convolutional network, extracting the three-dimensional features of each part, and obtaining a time flow three-dimensional feature graph;

[0095] Based on the flow dispersion graph, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature graph; the flow dispersion feature graph is a two-dimensional graph;

[0096] Among them, the size of the three-dimensional convolution kernel in the three-dimensional convolution network and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map; the size of the three-dimensional convolution kernel corresponding to the flow discrete map and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map.

[0097] Among them, the size of the three-dimensional convolution kernel in the three-dimensional convolution network is 2*2*2, the number of convolutions is 3 times, and the size of the three-dimensional convolution kernel corresponding to the flow discrete map is 4*n*4, the number of convolutions is 1 time, and the obtained flow discrete feature map and the time flow three-dimensional feature map have the same length and width, and the same size, and the n represents the width of the flow discrete feature map.

[0098] The sizes of the flow dispersion graph, the time-flow three-dimensional graph and the flow curve graph are made the same by padding with zeros.

[0099] The three-dimensional time flow graph, the flow curve characteristic graph and the flow discrete characteristic graph are superimposed to obtain a flow fusion characteristic graph.

[0100] The flow discrete characteristic diagram and the flow curve characteristic diagram are superimposed on the right side of the flow curve characteristic diagram, adjacent to the end of the curve.

[0101] Optionally, based on the flow dispersion map, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature map, including:

[0102] Obtaining a three-dimensional convolution kernel; the width of the three-dimensional convolution kernel is equal to the width of the flow dispersion map;

[0103] The three-dimensional convolution kernel moves sequentially in the length and channel direction of the flow discrete graph with a step size of 1; the three-dimensional convolution kernel moves sequentially from early to late in time.

[0104] Optionally, the detecting, based on the two-dimensional traffic graph, the location where the user does not use traffic, and obtaining the traffic disconnection time point includes:

[0105] A two-dimensional convolution kernel is obtained; the width of the two-dimensional convolution kernel is equal to the width of the two-dimensional flow map.

[0106] In this embodiment, the length of the two-dimensional convolution kernel is 2.

[0107] On the two-dimensional flow map, the two-dimensional convolution kernel is convolved from left to right with a step size of 1 to extract features and obtain a first convolution feature vector;

[0108] The first convolution feature vector is input into the first neural network to detect whether there is flow at a time point, and obtain a flow disconnected time point.

[0109] Optionally, the two-dimensional flow graph is segmented based on the flow unconnected time point to obtain a plurality of segmented flow graphs, including:

[0110] Search from left to right in the two-dimensional flow graph, detect the time point when the flow is not connected, cut it, and obtain the first local segmentation flow graph and the first segmentation flow graph to be detected; the time point corresponding to the first local segmentation flow graph is earlier than the time point corresponding to the first segmentation flow graph to be detected.

[0111] Search from left to right in the first segmented traffic graph to be detected, and if a time point is detected that is not a time point when the traffic is not connected, cut it to obtain a second segmented traffic graph to be detected; the starting time point of the second segmented traffic graph to be detected is the time point when the user connects to the traffic;

[0112] Search from left to right in the second segmented flow graph to be detected, detect the time point when the flow is not connected, and cut it to obtain the second local segmented flow graph and the third segmented flow graph to be detected;

[0113] Multiple searches correspond to obtaining multiple segmentation flow graphs.

[0114] The regions at time points where no discrete points exist are cut and deleted, thereby obtaining a plurality of segmented flow graphs retained after cutting.

[0115] Optionally, one segmented flow graph corresponds to one time point.

[0116] Optionally, the user discrimination network is a convolutional neural network.

[0117] The user discrimination network is a convolutional neural network (CNN), which is a user discrimination network trained according to labeled user discrimination categories.

[0118] Example 2

[0119] Based on the above-mentioned flow user data analysis method, an embodiment of the present invention further provides a flow user data analysis system, which includes an acquisition module, a curve module, a curve feature module, a fusion module and a category module.

[0120] The acquisition module is used to obtain the user traffic usage time point and the corresponding user traffic usage; the user traffic usage time point represents the time point of the traffic used by the user; the user traffic usage represents the traffic used by the user at the user traffic usage time point.

[0121] The curve module is used to construct multiple curves based on the user traffic usage time points and the corresponding user traffic usage to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; the traffic dispersion graph represents an image of the user traffic usage marked at multiple user traffic usage time points.

[0122] The curve feature module is used to project the features of multiple curves based on the flow curve diagram to obtain a flow curve feature diagram.

[0123] The fusion module is used to perform three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time points and the corresponding user traffic usage to obtain a traffic fusion characteristic graph.

[0124] The category module is used to input the traffic fusion feature map into the user discrimination network, discriminate the user's state, and obtain the user discrimination category; the user discrimination category indicates that the user is classified according to the adaptive traffic.

Claims

1. A method for analyzing traffic user data, characterized in that: include: Obtain the user traffic usage time point and the corresponding user traffic usage; The user traffic usage time point represents the time point of traffic usage by the user; The user traffic usage refers to the traffic used by the user at the time point of user traffic usage; Based on the user traffic usage time points and the corresponding user traffic usage, multiple curves are constructed to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; The traffic dispersion graph represents an image in which the traffic usage of the user is marked at multiple user traffic usage time points; Based on the flow curve diagram, projecting the characteristics of multiple curves to obtain a flow curve characteristic diagram; Based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage, a three-dimensional convolution is performed to obtain a traffic fusion characteristic graph; Inputting the traffic fusion feature graph into a user discrimination network to discriminate the user's state and obtain a user discrimination category; The user identification category indicates that users are classified according to the traffic they are suitable for.

2. A method for analyzing traffic user data according to claim 1, characterized in that: The method of constructing multiple curves based on the user traffic usage time points and the corresponding user traffic usage to obtain a traffic curve graph and a traffic dispersion graph includes: The user traffic usage time point is used as the horizontal axis, and the user traffic usage is used as the vertical axis to obtain a two-dimensional traffic graph; Based on the two-dimensional traffic graph, detect the location where the user does not use the traffic and obtain the time point when the traffic is not connected; Based on the flow unconnected time point, the two-dimensional flow graph is segmented to obtain a plurality of segmented flow graphs; Arrange multiple segmented flow graphs in order from early to late to obtain a flow dispersion graph; Fit the points in the segmented flow diagram to obtain a fitting curve; The fitting curve is plotted as an image to obtain a two-dimensional flow curve graph; a plurality of segmented flow graphs are correspondingly obtained to obtain a plurality of two-dimensional flow curve graphs; Arrange the plurality of two-dimensional flow curve graphs in sequence from early to late to obtain a flow curve graph.

3. A method for analyzing traffic user data according to claim 1, characterized in that: The method of projecting the characteristics of multiple curves based on the flow curve diagram to obtain a flow curve characteristic diagram includes: Taking the length as the dividing point, the side of the flow curve graph is divided multiple times to obtain multiple discrete time graphs; the width of the discrete time graph is equal to the width of the flow curve graph; the length of the discrete time graph is equal to the number of channels of the flow curve graph; Convolving the discrete moment graph to extract features and obtain a moment curve feature; multiple discrete moment graphs correspond to multiple moment curve features; The multiple time curve features are input into the time convolution network in sequence from early to late according to the time points to obtain a flow curve feature graph.

4. A method for analyzing traffic user data according to claim 1, characterized in that: The method of performing three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage to obtain a traffic fusion characteristic graph includes: Divide the user traffic usage time points into 24-hour time intervals to obtain multiple user interval time periods and corresponding user interval traffic; The user interval time period is used as the horizontal axis and the corresponding user interval flow rate is used as the vertical axis to draw an image to obtain a user interval two-dimensional image; multiple user interval time periods are corresponding to multiple user interval two-dimensional graphs; Multiple user interval two-dimensional graphs are superimposed according to time points from morning to night to obtain a three-dimensional time flow graph; Inputting the time flow three-dimensional graph into a three-dimensional convolutional network, extracting the three-dimensional features of each part, and obtaining a time flow three-dimensional feature graph; Based on the flow dispersion graph, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature graph; the flow dispersion feature graph is a two-dimensional graph; The size of the three-dimensional convolution kernel in the three-dimensional convolution network and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map; the size of the three-dimensional convolution kernel corresponding to the flow discrete map and the number of kernel convolutions make the size of the time flow three-dimensional feature map the same as the size of the flow curve feature map; The three-dimensional time flow graph, the flow curve characteristic graph and the flow discrete characteristic graph are superimposed to obtain a flow fusion characteristic graph; The flow discrete characteristic diagram and the flow curve characteristic diagram are superimposed on the right side of the flow curve characteristic diagram, adjacent to the end of the curve.

5. A method for analyzing traffic user data according to claim 2, characterized in that: Based on the flow dispersion graph, a three-dimensional convolution kernel is used to perform convolution in sequence from early to late time points to obtain a flow dispersion feature graph, including: Obtaining a three-dimensional convolution kernel; the width of the three-dimensional convolution kernel is equal to the width of the flow dispersion map; The three-dimensional convolution kernel moves sequentially in the length and channel direction of the flow discrete graph with a step size of 1; the three-dimensional convolution kernel moves sequentially from early to late in time.

6. A method for analyzing traffic user data according to claim 2, characterized in that: The method of detecting the location where the user does not use the traffic based on the two-dimensional traffic graph and obtaining the traffic disconnection time point includes: Obtaining a two-dimensional convolution kernel; the width of the two-dimensional convolution kernel is equal to the width of the two-dimensional flow map; On the two-dimensional flow map, the two-dimensional convolution kernel is convolved from left to right with a step size of 1 to extract features and obtain a first convolution feature vector; The first convolution feature vector is input into the first neural network to detect whether there is flow at a time point, and obtain a flow disconnected time point.

7. A method for analyzing traffic user data according to claim 2, characterized in that: The two-dimensional flow graph is segmented based on the flow unconnected time point to obtain a plurality of segmented flow graphs, including: Search from left to right in the two-dimensional flow graph, detect the time point when the flow is not connected, and cut it to obtain a first local segmented flow graph and a first segmented flow graph to be detected; the time point corresponding to the first local segmented flow graph is earlier than the time point corresponding to the first segmented flow graph to be detected; Search from left to right in the first segmented traffic graph to be detected, and if a time point is detected that is not a time point when the traffic is not connected, cut it to obtain a second segmented traffic graph to be detected; the starting time point of the second segmented traffic graph to be detected is the time point when the user connects to the traffic; Search from left to right in the second segmented flow graph to be detected, detect the time point when the flow is not connected, and cut it to obtain the second local segmented flow graph and the third segmented flow graph to be detected; Multiple searches are performed to obtain multiple local segmentation flow graphs; the multiple local segmentation flow graphs include a first local segmentation flow graph and a second local segmentation flow graph; The local segmentation flow graph is padded with 0 at both ends to obtain a segmentation flow graph.

8. A method for analyzing traffic user data according to claim 7, characterized in that: A segmented flow graph corresponds to a time point.

9. A method for analyzing traffic user data according to claim 3, characterized in that: The user discrimination network is a convolutional neural network.

10. A flow user data analysis system, characterized in that: include: Acquisition module: obtains the user traffic usage time point and the corresponding user traffic usage; The user traffic usage time point represents the time point of traffic usage by the user; The user traffic usage refers to the traffic used by the user at the time point of user traffic usage; Curve module: based on the user traffic usage time points and the corresponding user traffic usage, multiple curves are constructed to obtain a traffic curve graph and a traffic dispersion graph; the traffic curve graph represents an image of smooth changes in traffic at multiple user traffic usage time points; The traffic dispersion graph represents an image in which the traffic usage of the user is marked at multiple user traffic usage time points; Curve feature module: based on the flow curve diagram, projecting the features of multiple curves to obtain a flow curve feature diagram; Fusion module: performing three-dimensional convolution based on the traffic dispersion graph, the traffic curve characteristic graph, the user traffic usage time point and the corresponding user traffic usage to obtain a traffic fusion characteristic graph; Category module: input the traffic fusion feature map into the user discrimination network, discriminate the user's state, and obtain the user discrimination category; The user identification category indicates that users are classified according to the traffic they are suitable for.

Images