E-SIM user portrait generation method and system

Abstract

The present invention discloses an E‑SIM user portrait generation method and system. Each column of the traffic feature graph corresponds to the relationship between traffic data corresponding to multiple adjacent traffic times. The signal strength is converted into a two-dimensional matrix according to the meaning assigned to the column. Cluster clusters are constructed, the traffic data is divided into multiple parts, and the columns of corresponding traffic time points are marked. The corresponding traffic time points of the same cluster cluster and the surrounding features are extracted with circles, so that the features related to the same cluster cluster and the features with correlation relationships around can be obtained. The traffic data is classified according to the signal strength and filled into the matrix. Thereby achieving a more accurate acquisition of enhanced traffic change characteristics of the same cluster cluster, and the technical effect of being able to simultaneously distinguish the relationship between traffic time points, traffic data and signal strength and the user portrait of the E‑SIM card.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an E-SIM user portrait generation method and system. Background Art

[0002] At present, E-SIM (embedded SIM card) technology provides new possibilities for user portrait generation. Through E-SIM integrated in the device, user data can be collected and analyzed more effectively, thereby building a more accurate user portrait. Analysis of user behavior can also be obtained by analyzing the traffic data of users accessing the Internet. However, when generating user portraits, it is necessary to specifically extract the characteristics of users' traffic acquisition, but it is difficult to find the particularity and regularity of users' traffic usage. Summary of the invention

[0003] The purpose of the present invention is to provide an E-SIM user portrait generation method and system to solve the above-mentioned problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a method for generating an E-SIM user portrait, including:

[0005] Obtaining a flow time point, corresponding flow data and signal strength; the flow time point indicates a time point when a user uses flow; the flow data indicates the flow used at a flow time point; the signal strength indicates the strength of a signal at a flow time point;

[0006] Clustering the traffic data to obtain multiple traffic data clusters;

[0007] Based on a plurality of traffic data clusters, mapping the features of the traffic data to a plurality of feature spaces to obtain a first user feature graph;

[0008] Based on the first user characteristic graph, mapping the traffic data and the signal strength to a three-dimensional feature space to obtain a user characteristic graph;

[0009] The user feature map is stored in a database to form a user portrait.

[0010] Optionally, the mapping of the features of the traffic data to multiple feature spaces based on the multiple traffic data clusters to obtain a first user feature graph includes:

[0011] The flow data is subjected to feature extraction through a two-dimensional convolution kernel, and one-dimensional is converted into two-dimensional to obtain a flow feature map;

[0012] In the traffic feature graph, mark the time point corresponding to the traffic data cluster;

[0013] Based on the traffic characteristic graph, regional features are divided by circles to obtain a first user characteristic graph.

[0014] Optionally, the step of dividing the regional features by circles based on the traffic feature graph to obtain a first user feature graph includes:

[0015] Based on the traffic characteristic graph, find the row corresponding to the time point corresponding to the traffic data cluster and mark it to obtain the traffic characteristic point;

[0016] The flow characteristic point corresponding to the row marked in the flow characteristic graph is used as the center of the circle and the fixed length is used as the radius to construct a circle to obtain the flow characteristic circle;

[0017] Extract the features in the flow characteristic circle in the flow characteristic graph to obtain the first circle feature;

[0018] Based on the first circle feature and the traffic feature graph, the traffic feature points are superimposed to obtain the first user feature.

[0019] Optionally, based on the traffic characteristic graph, finding and marking the row corresponding to the time point corresponding to the traffic data cluster to obtain the traffic characteristic point includes:

[0020] A set of multiple traffic time points associated with the rows of the traffic characteristic graph is constructed to obtain a characteristic time point set; multiple rows of the traffic characteristic graph correspond to multiple characteristic time point sets;

[0021] According to the traffic data clustering cluster, the traffic time point corresponding to the traffic data is found to obtain a cluster time point set; multiple traffic data clustering clusters correspond to multiple cluster time point sets;

[0022] If the characteristic time point set includes a traffic time point in the cluster time point set, marking the row corresponding to the characteristic time point set as the corresponding traffic data cluster;

[0023] Using the row corresponding to the traffic data cluster in the traffic feature graph as the horizontal coordinate of the traffic feature point;

[0024] The column with half the number of columns of the flow characteristic graph is used as the ordinate of the flow characteristic point.

[0025] Optionally, mapping the traffic data and the signal strength to a three-dimensional feature space based on the first user characteristic graph to obtain a user characteristic graph includes:

[0026] Based on the signal strength and signal intensity, mapping features to obtain a three-dimensional feature matrix;

[0027] Performing three-dimensional convolution on the three-dimensional feature matrix, mapping features according to the size of the first user feature map, to obtain a second user feature map;

[0028] The first user characteristic map and the second user characteristic map are superimposed on the channel to obtain a user characteristic map.

[0029] Optionally, the obtaining of the first user feature by superimposing the first circle feature and the traffic feature graph according to traffic feature points includes:

[0030] The first circle features corresponding to the same flow data cluster are sequentially formed into a circle three-dimensional feature map in the order of time points; each channel of the circle three-dimensional feature map corresponds to a flow time point; the length of the circle three-dimensional feature map is equal to the number of flow time points;

[0031] If a channel of the circular three-dimensional feature map corresponds to a traffic time point that is not the same traffic data cluster, all values ​​in the channel of the circular three-dimensional feature map are 0;

[0032] Multiple traffic data clusters are clustered to obtain multiple circular three-dimensional feature maps;

[0033] Inputting the value of each channel in the circular three-dimensional feature map into the temporal convolutional network in sequence to obtain a circular temporal feature map;

[0034] Multiple circular three-dimensional feature maps correspond to multiple circular time feature maps;

[0035] Based on the traffic characteristic graph and a plurality of circle time characteristic graphs, a first user characteristic is obtained.

[0036] Optionally, obtaining the first user feature based on the traffic feature graph and the plurality of circle time feature graphs includes:

[0037] Convolving the flow characteristic graph to obtain a first convolution feature; the size of the first convolution feature is equal to the size of the circle time characteristic graph;

[0038] The traffic feature points are matched, and the first convolution feature and a plurality of circular time feature graphs are superimposed to obtain a first user feature.

[0039] Optionally, mapping features based on the signal strength and the signal intensity to obtain a three-dimensional feature matrix includes:

[0040] According to the flow time, the signal strength is transformed two-dimensionally in sequence to obtain a signal matrix; the rows of the signal matrix represent time points; the columns of the signal matrix represent signal strengths;

[0041] The flow data is divided according to the signal matrix to obtain a flow matrix;

[0042] The flow matrix and the signal matrix are superimposed to obtain a three-dimensional feature matrix.

[0043] Optionally, extracting features within the flow feature circle in the flow feature graph to obtain a first circle feature includes:

[0044] Extract the features within the flow feature circle to obtain a first extracted flow feature;

[0045] The first extracted flow feature is convolved to obtain a first circle feature.

[0046] In a second aspect, an embodiment of the present invention provides an E-SIM user portrait generation system, including:

[0047] Acquisition module: acquires flow time point, corresponding flow data and signal strength; the flow time point indicates the time point when the user uses flow; the flow data indicates the flow used at the flow time point; the signal strength indicates the strength of the signal at the flow time point;

[0048] Clustering module: clustering the traffic data to obtain multiple traffic data clusters;

[0049] Feature extraction module: based on multiple traffic data clusters, mapping the features of the traffic data to multiple feature spaces to obtain a first user feature graph;

[0050] A three-dimensional feature extraction module: based on the first user feature map, mapping the traffic data and signal strength to a three-dimensional feature space to obtain a user feature map;

[0051] Generation module: storing the user feature map in a database to form a user portrait.

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

[0053] An embodiment of the present invention also provides an E-SIM user portrait generation method and system, the method comprising: obtaining a traffic time point, corresponding traffic data and signal strength; the traffic time point indicates a time point when a user uses traffic; the traffic data indicates the traffic used at a traffic time point; the signal strength indicates the strength of the signal at the traffic time point; clustering the traffic data to obtain a plurality of traffic data clustering clusters; based on the plurality of traffic data clustering clusters, mapping the features of the traffic data to a plurality of feature spaces to obtain a first user feature graph; based on the first user feature graph, mapping the traffic data and signal strength to a three-dimensional feature space to obtain a user feature graph; storing the user feature graph in a database to form a user portrait.

[0054] The present invention uses a two-dimensional convolution kernel to convolve the flow data, thereby converting it into a two-dimensional feature map, that is, a flow feature map. Each column of the flow feature map corresponds to the relationship between the flow data corresponding to a plurality of adjacent flow times. The signal strength is converted into a two-dimensional matrix according to the meaning assigned to the column. A cluster is constructed, the flow data is divided into multiple parts, and the columns of the corresponding flow time points are marked. The corresponding flow time points of the same cluster and the surrounding features are extracted with circles, so that the features of the same cluster and the features with an associated relationship around can be obtained. These features are input into a time convolution network, so that the change of flow data during the time change process can be extracted, and the user portrait of the time change of the user's flow usage can be more accurately obtained. The flow data is classified according to the signal strength and filled into the matrix. The flow data and the signal strength form the same discrimination state, and the relationship between the flow time point, the flow data and the signal strength is jointly judged through three-dimensional detection, so as to achieve a more accurate acquisition of the enhanced flow change characteristics of the same cluster, and the technical effect of being able to simultaneously discriminate the relationship between the flow time point, the flow data and the signal strength. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a flow chart of a method for generating an E-SIM user portrait 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 generating an E-SIM user portrait, the method comprising:

[0059] S101: Obtain the flow time point, the corresponding flow data and the signal strength; the flow time point indicates the time point when the user uses the flow; the flow data indicates the flow used at the flow time point; the signal strength indicates the strength of the signal at the flow time point.

[0060] The traffic data is a one-dimensional vector arranged in sequence from early to late according to traffic time points.

[0061] S102: Clustering the traffic data to obtain a plurality of traffic data clusters.

[0062] In this embodiment, the k-means algorithm is used for clustering.

[0063] S103: Based on a plurality of traffic data clusters, mapping the features of the traffic data to a plurality of feature spaces to obtain a first user feature graph;

[0064] S104: Based on the first user characteristic graph, map the traffic data and the signal strength to a three-dimensional characteristic space to obtain a user characteristic graph;

[0065] S105: Storing the user feature map in a database to form a user portrait.

[0066] The user characteristic graph is used as a user portrait to facilitate subsequent judgment and detection.

[0067] Optionally, the mapping of the features of the traffic data to multiple feature spaces based on the multiple traffic data clusters to obtain a first user feature graph includes:

[0068] The flow data is subjected to feature extraction through a two-dimensional convolution kernel, and one-dimensional is converted into two-dimensional to obtain a flow feature map.

[0069] Among them, because the flow data is a one-dimensional vector, convolution is performed through a two-dimensional convolution kernel, and the flow data is expanded to obtain a two-dimensional flow characteristic map.

[0070] In the traffic feature graph, the time points corresponding to the traffic data clusters are marked.

[0071] Among them, because the traffic characteristic graph is the state of traffic data changes at multiple time points obtained during convolution, in the traffic characteristic graph, one value corresponds to multiple time points.

[0072] Based on the traffic characteristic graph, the marked traffic data are clustered and detected by dividing the regional features with circles to obtain a first user characteristic graph.

[0073] Optionally, the step of dividing the regional features by circles based on the traffic feature graph to obtain a first user feature graph includes:

[0074] Based on the traffic characteristic graph, find the row corresponding to the time point corresponding to the traffic data cluster and mark it to obtain the traffic characteristic point;

[0075] The flow characteristic point corresponding to the row marked in the flow characteristic graph is used as the center of the circle and a fixed length is used as the radius to construct a circle to obtain the flow characteristic circle.

[0076] In this embodiment, the fixed length is half of the width of the flow characteristic graph.

[0077] Extract the features in the flow characteristic circle in the flow characteristic graph to obtain the first circle feature;

[0078] Based on the first circle feature and the traffic feature graph, the traffic feature points are superimposed to obtain the first user feature.

[0079] Optionally, based on the traffic characteristic graph, finding and marking the row corresponding to the time point corresponding to the traffic data cluster to obtain the traffic characteristic point includes:

[0080] A set of multiple traffic time points associated with the rows of the traffic feature graph is constructed to obtain a characteristic time point set.

[0081] Among them, if a 2*2 two-dimensional convolution kernel is used to convolve the traffic data once, then the row corresponding to the traffic characteristic graph is associated with two adjacent traffic time points, such as the first row of the traffic characteristic graph is associated with the traffic data corresponding to the first and second traffic time points of the traffic data; if a 3*3 two-dimensional convolution kernel is used to convolve the traffic data once, then the row corresponding to the traffic characteristic graph is associated with three adjacent traffic time points, such as the first row of the traffic characteristic graph is associated with the traffic data corresponding to the first, second and third traffic time points of the traffic data.

[0082] According to the traffic data clustering cluster, the traffic time point corresponding to the traffic data is found to obtain a cluster time point set; multiple traffic data clustering clusters correspond to multiple cluster time point sets;

[0083] If the characteristic time point set includes a traffic time point in the cluster time point set, marking the row corresponding to the characteristic time point set as the corresponding traffic data cluster;

[0084] The row corresponding to the traffic data cluster in the traffic feature graph is used as the horizontal coordinate of the traffic feature point.

[0085] Among them, one or more rows in the traffic feature graph correspond to a traffic data cluster. That is, the one or more traffic feature points correspond to a traffic data cluster. Each traffic feature point is marked with a corresponding traffic data cluster.

[0086] The column with half the number of columns of the flow characteristic graph is used as the ordinate of the flow characteristic point.

[0087] Among them, because a column of the traffic feature graph reflects the characteristics of the traffic data at a traffic time point, the half position is used as the vertical coordinate of the traffic feature point, so the circle designed later can include the above-mentioned column and multiple surrounding features for detection.

[0088] Optionally, mapping the traffic data and the signal strength to a three-dimensional feature space based on the first user characteristic graph to obtain a user characteristic graph includes:

[0089] Based on the signal strength and signal intensity, features are mapped to obtain a three-dimensional feature matrix.

[0090] The three-dimensional feature matrix is ​​subjected to three-dimensional convolution, and features are mapped according to the size of the first user feature map to obtain a second user feature map.

[0091] Among them, a three-dimensional convolutional neural network (3D Convolutional Neural Networks, CNN) is used for convolution, and the size of the three-dimensional convolution kernel and the number of convolutions in the three-dimensional convolutional neural network (3D Convolutional Neural Networks, CNN) are designed according to the size of the first user feature map, so that the second user feature map is the same size as the first user feature map.

[0092] The length of the second user characteristic graph is equal to the length of the first user characteristic graph, and the width of the second user characteristic graph is equal to the width of the first user characteristic graph.

[0093] The first user characteristic map and the second user characteristic map are superimposed on the channel to obtain a user characteristic map.

[0094] Optionally, the obtaining of the first user feature by superimposing the first circle feature and the traffic feature graph according to traffic feature points includes:

[0095] The first circle features corresponding to the same traffic data cluster are sequentially constructed into a circular three-dimensional feature map in the order of time points; each channel of the circular three-dimensional feature map corresponds to a traffic time point; and the length of the circular three-dimensional feature map is equal to the number of traffic time points.

[0096] The number of the first circle features is equal to the number of flow feature points.

[0097] If a channel of the circular three-dimensional feature map corresponds to a traffic time point that is not the same traffic data cluster, all values ​​in the channel of the circular three-dimensional feature map are zero.

[0098] The circular three-dimensional feature map includes the first circle feature corresponding to the same traffic data cluster and a feature map of all zeros that does not correspond to the same traffic data cluster.

[0099] Multiple traffic data clusters correspond to multiple circular three-dimensional feature maps.

[0100] Inputting the value of each channel in the circular three-dimensional feature map into the temporal convolutional network in sequence to obtain a circular temporal feature map;

[0101] Multiple circular three-dimensional feature maps correspond to multiple circular time feature maps.

[0102] The traffic feature points are matched, and the first convolution feature and the first circle feature are superimposed to obtain a first user feature.

[0103] Optionally, obtaining the first user feature based on the traffic feature graph and the plurality of circle time feature graphs includes:

[0104] Convolving the flow characteristic graph to obtain a first convolution feature; the size of the first convolution feature is equal to the size of the circle time characteristic graph;

[0105] The traffic feature points are matched, and the first convolution feature and a plurality of circular time feature graphs are superimposed to obtain a first user feature.

[0106] The number of channels of the first user feature is equal to the sum of the number of channels of the first convolution feature and the number of channels of the first circle feature.

[0107] After the first circle feature is superimposed on the first convolution feature, the boundary of the first circle feature is padded with zeros, so that the first circle feature after the padded zeros has the same size as the first convolution feature.

[0108] Optionally, mapping features based on the signal strength and the signal intensity to obtain a three-dimensional feature matrix includes:

[0109] According to the flow time, the signal strength is transformed two-dimensionally in sequence to obtain a signal matrix; the rows of the signal matrix represent time points; the columns of the signal matrix represent signal strengths;

[0110] The flow data is divided according to the signal matrix to obtain a flow matrix.

[0111] Among them, the flow limit value is obtained; the flow limit value is the specified maximum flow that the user can use at the flow time point; the flow limit value is divided by the signal strength category; the signal strength category represents the category for classifying the signal strength. For example, if the signal strength is divided into 5 categories from weak to strong, the flow limit value is divided by 5 to obtain 5 flow ranges. Then the number of columns of the flow matrix is ​​5, and the number of rows is equal to the number of flow time points. Fill the flow data belonging to the flow range into the flow matrix.

[0112] The flow matrix and the signal matrix are superimposed to obtain a three-dimensional feature matrix.

[0113] Among them, the two two-dimensional matrices are superimposed into a three-dimensional matrix, the number of rows and columns are equal to the corresponding number of rows and columns of the traffic matrix and the signal matrix, and the number of pages is 2.

[0114] Optionally, extracting features within the flow feature circle in the flow feature graph to obtain a first circle feature includes:

[0115] The features within the flow feature circle are extracted to obtain the first extracted flow feature.

[0116] Among them, the first extracted flow feature is a feature after zero padding; the length of the first extracted flow feature is equal to twice the fixed length; the width of the first extracted flow feature is equal to twice the fixed length.

[0117] The first extracted flow feature is convolved to obtain a first circle feature.

[0118] Example 2

[0119] Based on the above-mentioned E-SIM user portrait generation method, an embodiment of the present invention further provides an E-SIM user portrait generation system, which includes an acquisition module, a clustering module, a feature extraction module, a three-dimensional feature extraction module and a generation module.

[0120] The acquisition module is used to obtain the flow time point, the corresponding flow data and signal strength; the flow time point indicates the time point when the user uses the flow; the flow data indicates the flow used at the flow time point; the signal strength indicates the strength of the signal at the flow time point.

[0121] The clustering module is used to cluster the traffic data to obtain multiple traffic data clusters.

[0122] The feature extraction module is used to map the features of the traffic data to multiple feature spaces based on multiple traffic data clusters to obtain a first user feature graph.

[0123] The three-dimensional feature extraction module is used to map the traffic data and signal strength to a three-dimensional feature space based on the first user feature map to obtain a user feature map.

[0124] The generation module is used to store the user feature map in a database to form a user portrait.

Claims

1. A method for generating an E-SIM user profile, characterized in that: include: Obtain traffic time points, corresponding traffic data and signal strength; The traffic time point indicates the time point when the user uses the traffic; The flow data represents the flow used at a flow time point; The signal strength represents the strength of the signal at the flow time point; Clustering the traffic data to obtain multiple traffic data clusters; Based on a plurality of traffic data clusters, mapping the features of the traffic data to a plurality of feature spaces to obtain a first user feature graph; Based on the first user characteristic graph, mapping the traffic data and the signal strength to a three-dimensional feature space to obtain a user characteristic graph; Storing the user feature map in a database to form a user portrait; The method of mapping the characteristics of the traffic data to multiple feature spaces based on the multiple traffic data clusters to obtain a first user feature graph includes: The flow data is subjected to feature extraction through a two-dimensional convolution kernel, and one-dimensional is converted into two-dimensional to obtain a flow feature graph; in the flow feature graph, the time points corresponding to the flow data clusters are marked; Based on the traffic characteristic graph, the regional characteristics are divided by circles to obtain a first user characteristic graph; The method of dividing the regional features by circles based on the traffic characteristic graph to obtain a first user characteristic graph includes: Based on the traffic characteristic graph, find the row corresponding to the time point corresponding to the traffic data cluster and mark it to obtain the traffic characteristic point; use the traffic characteristic point corresponding to the row marked in the traffic characteristic graph as the center of the circle and a fixed length as the radius to construct a circle to obtain the traffic characteristic circle; extract the features in the traffic characteristic circle from the traffic characteristic graph to obtain the first circle feature; based on the first circle feature and the traffic characteristic graph, superimpose them according to the traffic characteristic points to obtain the first user feature; The method of finding and marking the row corresponding to the time point corresponding to the traffic data cluster based on the traffic characteristic graph to obtain the traffic characteristic point includes: Construct a set of multiple traffic time points associated with the rows of the traffic feature graph to obtain a characteristic time point set; obtain multiple characteristic time point sets corresponding to multiple rows of the traffic feature graph; find the traffic time points corresponding to the traffic data according to the traffic data clustering cluster to obtain a clustering cluster time point set; obtain multiple clustering cluster time point sets corresponding to multiple traffic data clustering clusters; if the characteristic time point set contains the traffic time points in the clustering cluster time point set, mark the rows corresponding to the characteristic time point set as the corresponding traffic data clustering cluster; use the rows corresponding to the traffic data clustering clusters in the traffic feature graph as the horizontal coordinates of the traffic feature points; use half of the number of columns of the traffic feature graph as the vertical coordinates of the traffic feature points; The mapping of the traffic data and the signal strength to a three-dimensional feature space based on the first user characteristic graph to obtain a user characteristic graph includes: Based on the signal strength and signal intensity, mapping features to obtain a three-dimensional feature matrix; Performing three-dimensional convolution on the three-dimensional feature matrix, mapping features according to the size of the first user feature map, to obtain a second user feature map; Superimposing the first user characteristic map and the second user characteristic map on the channel to obtain a user characteristic map; The method of obtaining the first user feature based on the first circle feature and the traffic feature graph and superimposing the traffic feature points to obtain the first user feature includes: The first circle features corresponding to the same traffic data clustering cluster are sequentially used to form a circle three-dimensional feature map in the order of time points; each channel of the circle three-dimensional feature map corresponds to a traffic time point; the length of the circle three-dimensional feature map is equal to the number of traffic time points; wherein, if a channel of the circle three-dimensional feature map corresponds to a traffic time point that is not the same traffic data clustering cluster, the values ​​in a channel of the circle three-dimensional feature map are all 0; multiple traffic data clustering clusters correspond to multiple circle three-dimensional feature maps; the value of each channel in the circle three-dimensional feature map is sequentially input into the temporal convolutional network to obtain a circle time feature map; multiple circle three-dimensional feature maps correspond to multiple circle time feature maps; based on the traffic feature map and multiple circle time feature maps, a first user feature is obtained.

2. The method for generating an E-SIM user portrait according to claim 1, characterized in that: The method of obtaining a first user feature based on the traffic feature graph and multiple circle time feature graphs includes: convolving the traffic feature graph to obtain a first convolution feature; the size of the first convolution feature is equal to the size of the circle time feature graph; matching traffic feature points, superimposing the first convolution feature and multiple circle time feature graphs, and obtaining a first user feature.

3. The method for generating an E-SIM user portrait according to claim 1, characterized in that: The mapping of features based on the signal strength and signal strength to obtain a three-dimensional feature matrix includes: performing two-dimensional transformation of the signal strength in turn according to the flow time to obtain a signal matrix; the rows of the signal matrix represent time points; the columns of the signal matrix represent signal strengths; dividing the flow data according to the signal matrix to obtain a flow matrix; The flow matrix and the signal matrix are superimposed to obtain a three-dimensional feature matrix.

4. The method for generating an E-SIM user portrait according to claim 1, characterized in that: The extracting the features in the flow feature circle in the flow feature graph to obtain the first circle feature includes: extracting the features in the flow feature circle to obtain the first extracted flow feature; and convolving the first extracted flow feature to obtain the first circle feature.

5. An E-SIM user portrait generation system, characterized in that: include: Acquisition module: obtains traffic time points, corresponding traffic data and signal strength; The traffic time point indicates the time point when the user uses the traffic; The flow data represents the flow used at a flow time point; The signal strength represents the strength of the signal at the flow time point; A clustering module: clustering the traffic data to obtain a plurality of traffic data clusters; a feature extraction module: mapping the features of the traffic data to a plurality of feature spaces based on the plurality of traffic data clusters to obtain a first user feature graph; A three-dimensional feature extraction module: based on the first user feature map, mapping the traffic data and signal strength to a three-dimensional feature space to obtain a user feature map; Generating module: storing the user feature map in a database to form a user portrait; The method of mapping the characteristics of the traffic data to multiple feature spaces based on the multiple traffic data clusters to obtain a first user feature graph includes: The flow data is subjected to feature extraction through a two-dimensional convolution kernel, and one-dimensional is converted into two-dimensional to obtain a flow feature graph; in the flow feature graph, the time points corresponding to the flow data clusters are marked; Based on the traffic characteristic graph, the regional characteristics are divided by circles to obtain a first user characteristic graph; The method of dividing the regional features by circles based on the traffic characteristic graph to obtain a first user characteristic graph includes: Based on the traffic characteristic graph, find the row corresponding to the time point corresponding to the traffic data cluster and mark it to obtain the traffic characteristic point; use the traffic characteristic point corresponding to the row marked in the traffic characteristic graph as the center of the circle and a fixed length as the radius to construct a circle to obtain the traffic characteristic circle; extract the features in the traffic characteristic circle from the traffic characteristic graph to obtain the first circle feature; based on the first circle feature and the traffic characteristic graph, superimpose them according to the traffic characteristic points to obtain the first user feature; The method of finding and marking the row corresponding to the time point corresponding to the traffic data cluster based on the traffic characteristic graph to obtain the traffic characteristic point includes: Construct a set of multiple traffic time points associated with the rows of the traffic feature graph to obtain a characteristic time point set; obtain multiple characteristic time point sets corresponding to multiple rows of the traffic feature graph; find the traffic time points corresponding to the traffic data according to the traffic data clustering cluster to obtain a clustering cluster time point set; obtain multiple clustering cluster time point sets corresponding to multiple traffic data clustering clusters; if the characteristic time point set contains the traffic time points in the clustering cluster time point set, mark the rows corresponding to the characteristic time point set as the corresponding traffic data clustering cluster; use the rows corresponding to the traffic data clustering clusters in the traffic feature graph as the horizontal coordinates of the traffic feature points; use half of the number of columns of the traffic feature graph as the vertical coordinates of the traffic feature points; The mapping of the traffic data and the signal strength to a three-dimensional feature space based on the first user characteristic graph to obtain a user characteristic graph includes: Based on the signal strength and signal intensity, mapping features to obtain a three-dimensional feature matrix; Performing three-dimensional convolution on the three-dimensional feature matrix, mapping features according to the size of the first user feature map, to obtain a second user feature map; Superimposing the first user characteristic map and the second user characteristic map on the channel to obtain a user characteristic map; The method of obtaining the first user feature based on the first circle feature and the traffic feature graph and superimposing the traffic feature points to obtain the first user feature includes: The first circle features corresponding to the same traffic data clustering cluster are sequentially used to form a circle three-dimensional feature map in the order of time points; each channel of the circle three-dimensional feature map corresponds to a traffic time point; the length of the circle three-dimensional feature map is equal to the number of traffic time points; wherein, if a channel of the circle three-dimensional feature map corresponds to a traffic time point that is not the same traffic data clustering cluster, the values ​​in a channel of the circle three-dimensional feature map are all 0; multiple traffic data clustering clusters correspond to multiple circle three-dimensional feature maps; the value of each channel in the circle three-dimensional feature map is sequentially input into the temporal convolutional network to obtain a circle time feature map; multiple circle three-dimensional feature maps correspond to multiple circle time feature maps; based on the traffic feature map and multiple circle time feature maps, a first user feature is obtained.

Images