CN113573242B - Identification method, device and equipment of re-networking user - Google Patents

Abstract

The invention provides a method, a device and equipment for identifying a re-networking user. The method comprises the following steps: acquiring behavior feature vectors of at least two users; the behavior feature vector records the occurrence time, the spatial position and the intensity representation information of the target behavior; according to the characteristic points of the behavior characteristic vector of each user determined in the space-time behavior characteristic cube, carrying out similarity analysis on the characteristic points of the first user and the characteristic points of the second user, and judging whether the second user is a re-network user of the first user; the time, the longitude of the space position and the latitude of the space position are used as coordinates of the space-time behavior feature cube, and the feature points are distributed in the space-time behavior feature cube according to the time and the space position of the target behavior. The identification method utilizes the time, the space position and the intensity representation information to construct the behavior characteristic vector of the user, compares the characteristic points of different users, and can simply and effectively identify the re-entry network user in the same operation network.

Description

Method, device and equipment for identifying re-entrant network users

technical field

The present invention relates to the field of communication technology, in particular to a method, device and equipment for identifying re-entry network users.

Background technique

Users who are or have used a card number of a certain operator purchase a card number belonging to the same operator in a short period of time to access the network, and the new number completely or partially replaces the old number. These users are re-entry users, and re-entry users occupy system card number resources , which increases the company's marketing costs and increases business risks, so it needs to be effectively identified and managed. However, since the re-entry mobile phone number and the original online mobile phone number are two different numbers, how to determine whether the two numbers belong to the same person is the key to identifying the re-entry number.

Contents of the invention

The purpose of the technical solution of the present invention is to provide a method, device and equipment for identifying re-entry users, which can simply and effectively identify re-entry users in the same operating network.

An embodiment of the present invention provides a method for identifying a re-entry user, which includes:

Obtaining behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior;

According to the feature points determined by the behavior feature vector of each user in the spatio-temporal behavior feature cube, the feature points of the first user and the feature points of the second user among at least two users are analyzed for similarity, and the second user is judged. Whether the user is a re-entry user of the first user;

Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior.

Optionally, the method for identifying re-entry users, wherein the acquiring behavioral feature vectors of at least two users includes:

Collect behavior data of each user; the behavior data includes time, space position and intensity representation information of different target behaviors;

Constructing the spatiotemporal behavior characteristic cube of each user according to the behavior data;

performing cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determining that the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is the behavior feature vector;

Deleting other feature points in the spatio-temporal behavior feature cube other than the feature points corresponding to the behavior feature vector.

Optionally, in the method for identifying re-entry users, wherein the cluster analysis is performed on the behavior data in the spatio-temporal behavior characteristic cube, and the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is determined to be Behavior feature vectors, including:

Slicing the spatio-temporal behavior feature cube according to the time dimension to form a plurality of slice data;

Clustering the behavioral data in each slice data to determine at least one clustering point;

The intensity representation information of the behavior data corresponding to each cluster point is compared with a preset intensity threshold, and the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is determined as the behavior feature vector.

Optionally, in the method for identifying re-entry users, the intensity indication information is expressed as the cumulative duration of the target behavior within a preset statistical period.

Optionally, the method for identifying re-entry users, wherein, before performing similarity analysis on the feature points of the first user and the feature points of the second user among the at least two users, the method further includes:

The behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user are deunited and standardized to obtain the normalized data vector of the first user and the normalized data vector of the second user ; Each normalized data vector corresponds to a feature point;

Wherein, the similarity analysis is performed on the feature points of the first user and the feature points of the second user among at least two users, including:

Perform similarity analysis on the feature points corresponding to the normalized data vector of the first user and the feature points corresponding to the normalized data vector of the second user to determine whether the second user is a re-entry user of the first user.

Optionally, in the method for identifying re-entry users, the similarity analysis is performed on the feature points of the first user and the feature points of the second user among at least two users, and it is judged whether the second user is the Re-entry users of the first user, including:

determining similar feature points between the feature points of the first user and the feature points of the second user;

When the ratio between the number of similar feature points and the number of feature points of the first user is greater than a first preset value, it is determined that the second user is a re-entry user of the first user.

Optionally, in the method for identifying re-entry users, the determining similar feature points between the feature points of the first user and the feature points of the second user includes:

Selecting the first feature point of the first user;

Calculating a second feature point of the second user with the shortest distance from the first feature point; wherein the target behavior of the behavior feature vector corresponding to the first feature point and the second feature point is the same;

Analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature point are similar feature points.

Optionally, the method for identifying re-entry users, wherein, analyzing the similarity value between the first feature point and the second feature point, and judging the difference between the first feature point and the second feature point Whether it is a similar feature point, including:

Acquiring a first weight value when the target behavior corresponding to the first feature point occurs within a preset time period, and acquiring a second weight value when the target behavior corresponding to the second feature point occurs within a preset time period;

determining a weight coefficient according to the first weight value and the second weight value;

calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point;

When it is determined that the similarity value is greater than a second preset value, the first feature point and the second feature are similar feature points.

Optionally, the method for identifying re-entry users, wherein, according to the first weight value and the second weight value, determining a weight coefficient includes:

calculating the ratio of the minimum value of the first weight value and the second weight value to the maximum value of the first weight value and the second weight value;

The ratio is determined as the weight coefficient.

Optionally, the method for identifying re-entry users, wherein, according to the weight coefficient and the distance between the first feature point and the second feature point, calculating the similarity value includes:

The similarity value is calculated according to the following formula:

Si=1-Di/Wi;

Wherein, Si is a similarity value; Di is a distance between the first feature point and the second feature point; Wi is the weight coefficient.

Optionally, the method for identifying re-entry users, wherein, before de-uniting and normalizing the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user , the method also includes:

In the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the time distribution dimension of the behavioral feature vector of the first user and the behavioral feature vector of the second user is determined to determine the time domain segmentation point;

Segment and assemble the spatiotemporal behavior characteristic cube of the first user and the spatiotemporal behavior characteristic cube of the second user, so that the spatiotemporal behavior characteristic cube of the first user and the second user's spatiotemporal behavior characteristic cube after segmentation and assembling The spatio-temporal behavior characteristic cube uses the time corresponding to the time domain segmentation point as the starting time point;

Wherein, the de-unit standardization transformation is performed on the behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user, including:

De-unit normalization is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube after segmentation and assembly.

Optionally, in the method for identifying re-entry users, wherein, in the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the behavioral feature vector of the first user and the behavioral feature vector of the second user Second, the time distribution dimension of the user's behavior feature vector, to determine the time domain segmentation point, including:

According to the time dimension, the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vectors in the spatio-temporal behavior feature cube of the second user respectively accumulate the intensity representation information corresponding to the same target behavior;

Draw a first intensity change curve of the behavior feature vector of the first user and draw a second intensity change curve of the behavior feature vector of the second user according to the accumulated maximum intensity information value corresponding to each target behavior;

The lowest point of the first intensity variation curve and the second intensity variation curve is selected as the time domain segmentation point.

Optionally, in the method for identifying re-entry users, wherein, the behavior feature vector in the first user's spatio-temporal behavior feature cube and the behavior feature vector in the second user's spatio-temporal behavior feature cube are de-united and standardized, include:

By means of a dispersion normalization method or a standard deviation normalization method, de-unit normalization conversion is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube respectively.

An embodiment of the present invention also provides an identification device for a re-entry user, which includes:

A vector acquisition module, configured to acquire behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior;

The comparison module is used to perform similarity analysis on the feature points of the first user and the feature points of the second user among at least two users according to the feature points determined in the spatio-temporal behavior feature cube by the behavior feature vector of each user, judging whether the second user is a re-entry user of the first user;

Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior.

An embodiment of the present invention also provides an identification device, which includes: a processor, a memory, and a program stored on the memory and operable on the processor. When the program is executed by the processor, the following steps are implemented: The method for identifying re-entry users described in any one of the preceding items.

An embodiment of the present invention also provides a readable storage medium, wherein a program is stored on the readable storage medium, and when the program is executed by a processor, the method for identifying re-entrant users as described in any one of the above items is implemented. step.

At least one of the above technical solutions of the present invention has the following beneficial effects:

Using the identification method for re-entry users described in the embodiment of the present invention, using time, spatial position and intensity representation information to construct user behavior feature vectors, and performing similarity analysis according to the feature points determined by the behavior feature vectors in the spatio-temporal behavior feature cube , to identify re-entry users, the identification method associates time, space position and intensity representation information, compared with the prior art, it can simply and effectively identify re-entry users in the same operating network.

Description of drawings

FIG. 1 is a schematic flow diagram of a method for identifying re-entry users according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of step S110 in FIG. 1;

3 is a schematic diagram of one of the spatiotemporal behavioral feature cubes;

FIG. 4 is a schematic flow chart of step S113 in FIG. 2;

Figure 5 is a schematic diagram of one of the time slices;

FIG. 6 is a schematic flow chart of step S120 in FIG. 1;

Fig. 7 is a schematic structural diagram of the space-time behavior characteristic cube after time domain segmentation and assembly;

FIG. 8 is a schematic flow diagram of an identification device for a re-entry user according to an embodiment of the present invention;

Fig. 9 is a schematic structural diagram of an identification device according to an embodiment of the present invention.

Detailed ways

In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will describe in detail with reference to the drawings and specific embodiments.

An embodiment of the present invention provides a method for identifying re-entry users, which utilizes the characteristics of high similarity in the behavior of the same user at a specific time and space, uses time, space position and intensity representation information to construct a user's behavior feature vector, and according to the behavior characteristics The feature points determined by the vector in the spatio-temporal behavior feature cube correlate time, space position and intensity representation information, and can easily and effectively identify re-entry users in the same operating network.

The method for identifying re-entry users described in one of the embodiments of the present invention, as shown in Figure 1, the method includes:

S110, acquiring behavior feature vectors of at least two users; the behavior feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior;

S120, according to the feature points determined by the behavior feature vector of each user in the spatio-temporal behavior feature cube, perform a similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, and determine the Whether the second user is a re-entry user of the first user;

Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior.

Using the method for identifying re-entry users described in the embodiments of the present invention, using time, spatial position and intensity representation information to construct user behavior feature vectors, and comparing the feature points determined by the behavior feature vectors of different users in the spatiotemporal behavior feature cube, Compared with using a single-dimensional model for user identification, the identification of re-entry users can effectively guarantee the accuracy of re-entry user identification; in addition, using this method for re-entry user identification correlates time, spatial location and intensity representation information Combined analysis can further effectively ensure the accuracy of re-entry user identification, and compared with the existing technology, each dimension is modeled separately for similarity analysis, and then the similarity analysis of each dimension is comprehensively analyzed to identify re-entry users. The identification method, the identification method described in the embodiment of the present invention, is simpler and easier to implement.

Optionally, as shown in FIG. 2, in step S110, at least two user behavior feature vectors are obtained, including:

S111, collect the behavior data of each user; the behavior data includes time, space position and intensity representation information of different target behaviors;

S112, constructing the spatio-temporal behavior characteristic cube of each user according to the behavior data; wherein the spatio-temporal behavior characteristic cube takes time, the longitude of the spatial position and the latitude of the spatial position as coordinates, and the behavior data is based on the time of the target behavior and spatial positions are distributed in the spatiotemporal behavioral feature cube;

S113. Perform cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determine that the corresponding behavior data whose intensity representation information is greater than a preset intensity threshold is the behavior feature vector;

S114. Delete other feature points in the spatio-temporal behavior feature cube other than the feature points corresponding to the behavior feature vector.

In step S111, the behavior data of each user is collected, including collecting four elements: time, spatial location, behavior and intensity representation information.

1) Time: With a preset period (for example, 10 minutes) as the period, collect the behavior data of the user within a specified period of 24 hours a day for a period of time (for example, two weeks);

2) Spatial location: including latitude and longitude information; among them, the location information of the base station within a specified time period is collected through signaling data, call and Internet access data, etc., and converted into latitude and longitude information to obtain the spatial location in the behavior data;

3) Behavior: Optionally, the collected behaviors may include: the user is in the power-on state but does not have a call or surf the Internet, for example, the user is on standby; the call behavior is distinguished by number segment, such as C_139; the application program APP behavior is distinguished by APP, etc. .

4) Intensity representation information: Optionally, the intensity representation information is expressed as the cumulative duration of the target behavior within the preset statistical period; it should be noted that the cumulative duration is the cumulative duration of the target behavior occurring in the target space within the preset statistical period . For example, the user's behavior intensity representation information is identified by the cumulative behavior duration (in seconds) of a location within a period of time (for example, within two weeks).

By collecting the behavior data of each user based on the above four elements, multiple groups of behavior data corresponding to different target behaviors, including target behavior, time, spatial position and intensity representation information, are obtained. For example, an example of collected behavioral data may be shown in Table 1 below:

Table 1

time longitude dimension Behavior strength 10:00—10:10 115"25'E Latitude 39"26'N user standby 500 10:10—10:20 115"25'E Latitude 39"26'N call behavior 40 10:20—10:30 115"26'E Latitude 39"27'N APP behavior 600 …

On the condition that the behavior data in the above form is obtained through step S111, in the embodiment of the present invention, in step S112, a spatio-temporal behavior characteristic cube is constructed using the obtained behavior data. Among them, in the spatio-temporal behavior feature cube, time, the longitude of the spatial position and the latitude of the spatial position are respectively the three-dimensional coordinates of the feature position, and each behavior data of the user is characterized by the three dimensions of longitude, latitude and time, Distributed in the spatiotemporal behavioral feature cube and represented by feature points, each behavioral data corresponds to a feature point.

An example of the spatio-temporal behavior feature cube can be shown in Figure 3. Different target behaviors are represented by different gray levels in Figure 3, which can be distinguished by color during specific implementation. According to Figure 3, through the spatio-temporal behavior feature cube, the distribution of different target behaviors in time and space can be clearly displayed.

It should be noted that, through the above method of constructing a spatio-temporal behavior feature cube, a corresponding spatio-temporal behavior feature cube can be constructed for each user.

Further, in the method for identifying re-entry users described in the embodiment of the present invention, after constructing the spatiotemporal behavior characteristic cube in step S112, cluster analysis is performed on the behavior data in the spatiotemporal behavior characteristic cube in step S113, and it is determined that the intensity indicates that the information is greater than The behavioral data corresponding to the preset intensity threshold is a behavioral feature vector, and the feature points corresponding to the behavioral feature vector determined in the spatio-temporal behavioral feature cube are reserved for subsequent similarity in the identification of re-entry users analyze.

Optionally, in step S113, performing cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determining that the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is a behavior feature vector, as shown in FIG. 4 ,include:

S1131. Slicing the spatio-temporal behavior feature cube according to the time dimension to form a plurality of slice data;

S1132, cluster the behavior data in each slice data, and determine at least one cluster point;

S1133. Compare the intensity representation information of the behavior data corresponding to each cluster point with a preset intensity threshold, and determine that the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is the behavior feature vector.

In step S1131, when slicing the spatio-temporal behavior feature cube according to the time dimension, the spatio-temporal behavior feature cube can be switched according to the preset duration of each interval according to the time dimension segment to form multiple slice data; for example, for all The constructed user's spatio-temporal behavior feature cube forms 48 slices of data according to the way of slice every half hour.

In step S1132, the behavioral data in each slice data is clustered. Optionally, for each slice data, a density-based clustering method (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) can be used. For clustering, each cluster takes its center point as a cluster point, and the cluster point as a feature identifier, which can reflect the average position and average intensity of a user's specific behavior at a specific time, so that data can be extracted as a target and build a user profile. Behavioral feature vector of .

For example, suppose a user takes the subway home from get off work at around 6:00 every day, and likes to play Douyin on the subway, then the simplified version of the sliced data that has undergone density clustering from 6:00 to 6:30 is shown in Figure 5, black The dots represent the user's standby behavior, the white dots represent the behavior of playing Douyin, and the size of the dots represents the intensity of the behavior.

Therefore, through the above step S1132, at least one cluster point in each slice data can be determined, and the cluster point can reflect the average position and average intensity of the user's specific behavior at a specific time.

On this basis, through step S1133, the intensity representation information of the behavior data corresponding to each cluster point is compared with the preset intensity threshold value, that is, the threshold value judgment is performed on the intensity representation information of each cluster point, and the intensity representation information is determined. The corresponding behavioral data whose information is greater than the preset intensity threshold is the behavioral feature vector, that is, it is determined to be valid data, saved, and constructed as the user's behavioral feature vector.

In one implementation of the embodiments of the present invention, as shown in FIG. 6, in step S120, before performing similarity analysis on the feature points of the first user and the feature points of the second user among the at least two users, the method further include:

S1101. Perform de-unit normalization transformation on the behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user, to obtain the normalized data vector of the first user and the normalized data vector of the second user Data vector; wherein each normalized data vector corresponds to a feature point;

Wherein, in step S120, the similarity analysis is performed on the feature points of the first user and the feature points of the second user among at least two users, including:

Perform similarity analysis on the feature points corresponding to the normalized data vector of the first user and the feature points corresponding to the normalized data vector of the second user to determine whether the second user is a re-entry user of the first user.

Specifically, through the above method, standardize the data of the user's behavior feature vector, remove the unit restriction of the data, and convert it into a dimensionless pure value, so that indicators of different units or magnitudes can be calculated and compared, so that they can be used for Subsequent similarity comparisons of feature points corresponding to standardized data vectors of different users.

In the embodiment of the present invention, in order to ensure the accuracy of data analysis, optionally, in step S1101, the behavior feature vector in the first user's spatio-temporal behavior feature cube and the behavior feature vector in the second user's spatio-temporal behavior feature cube are Before de-unit normalization conversion, the method also includes:

In the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the time distribution dimension of the behavioral feature vector of the first user and the behavioral feature vector of the second user is determined to determine the time domain segmentation point;

Segment and assemble the spatiotemporal behavior characteristic cube of the first user and the spatiotemporal behavior characteristic cube of the second user, so that the spatiotemporal behavior characteristic cube of the first user and the second user's spatiotemporal behavior characteristic cube after segmentation and assembling The spatio-temporal behavior characteristic cube uses the time corresponding to the time domain segmentation point as the starting time point;

Wherein, in step S1101, the behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user are deunited and standardized, including:

De-unit normalization is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube after segmentation and assembly.

Wherein, optionally, in the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user respectively, the time distribution dimension of the behavioral feature vector of the first user and the behavioral feature vector of the second user , to determine the time domain segmentation point, including:

According to the time dimension, the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vectors in the spatio-temporal behavior feature cube of the second user respectively accumulate the intensity representation information corresponding to the same target behavior;

Draw a first intensity change curve of the behavior feature vector of the first user and draw a second intensity change curve of the behavior feature vector of the second user according to the accumulated maximum intensity information value corresponding to each target behavior;

The lowest point of the first intensity variation curve and the second intensity variation curve is selected as the time domain segmentation point.

The first user is set as the original online user, the second user is the user to be matched, and the behavior feature vectors of the first user and the second user are extracted from the database, and the behavior features of the first user and the second user are Before the vector is de-united and normalized, the time-domain segmentation point is determined, and the time-domain segmentation is performed.

By determining the time-domain segmentation point and selecting the weakest point of user activity intensity, according to the weaker range of user activity intensity, the above-mentioned determined spatio-temporal behavior characteristic cube can be re-segmented and assembled for subsequent identification of re-entry users.

It should be noted that since the occurrence time of the target behavior recorded in the behavior feature vector is within the time range of one day, if the default 0-point time axis is used as the starting point for behavior feature extraction and analysis, it is easy to make a wrong judgment. This is because if the user has behaviors near 0 o'clock, such as the same behavior in the same place in the period of 23:30-24:00 and in the period of 00:00-00:30, but the actual interval is within one hour, at time In the feature space where the axis is 0, the time interval of the same behavior is more than 20 hours, so the interval of the data will lead to misjudgment.

Based on this, in the identification method described in the embodiment of the present invention, through the above-mentioned method of selecting the weakest point of user activity intensity, the selected weakest point of user activity intensity is used to re-switch and arrange the behavior feature vector according to the time domain. When the above-mentioned behavioral feature vectors are used to construct the spatio-temporal behavioral feature cube, the constructed spatio-temporal behavioral feature cube is re-segmented and assembled, and the determined time-domain segmentation point is used as the starting time point of behavioral feature vector distribution. For example, as shown in Figure 7, the time-domain segmentation point of the spatio-temporal behavior feature cube determined according to the behavior feature vectors of user A and user B is 2 points, then after time-domain segmentation and assembling of the spatio-temporal behavior feature cube, 2 The point is transformed into the time axis starting point.

In the embodiment of the present invention, by accumulating the intensity representation information corresponding to the same target behavior on the spatio-temporal behavior characteristic cube of the first user and the behavior characteristic cube of the second user according to the time dimension, according to each target behavior Corresponding to the accumulated maximum intensity information value, draw the first intensity change curve of the behavior feature vector of the first user, and draw the second intensity change curve of the behavior feature vector of the second user; select the first intensity The lowest point of the change curve and the second intensity change curve is the time domain cut point. That is, extract the behavior feature vectors of the two users to be matched, accumulate the user behavior intensity according to the time dimension, select the point where the maximum behavior intensity occurs as the starting point, draw the 24-hour change curve of the behavior intensity, and then select the two users The extremely low point of the corresponding behavior change curve is used as the cut-off point in the time domain.

In the recognition method described in the embodiment of the present invention, after the time-domain segmentation of the user's spatio-temporal behavior feature cube is performed, the behavior feature vector in each user's spatio-temporal behavior feature cube is further de-united and standardized. Specifically, the behavior feature vector in the spatio-temporal behavior feature cube of each user can be de-united and normalized through the dispersion standardization method or the standard deviation standardization method, so as to remove the unit limitation of the data in the behavior feature vector and transform it into a dimensionless The pure value of , which facilitates the calculation and comparison of indicators with different units or magnitudes.

In one implementation manner, the original behavioral feature vector is subjected to de-unit normalization conversion by using a deviation normalization method. The formula used in this conversion method can be:

X'=(X-min)/(max-min);

Among them, X' is the data after conversion, X is the data before conversion, max is the maximum value in the converted sample data, and min is the minimum value in the converted sample data.

In this way, the data of each dimension of the behavioral feature vector can be converted into data in the [0, 1] interval, thereby removing the unit limitation of different dimensional data in the behavioral feature vector.

For example, using this method, the occurrence time, longitude, and latitude in the behavioral feature vector can be denormalized and converted to obtain the behavioral feature vector shown in Table 2 below:

Table 2

time mean longitude average dimension Behavior average intensity 0.12 0.345 0.567 B 210 0.34 0.12 0.8 C 120 …

In another implementation manner, the original behavioral feature vector is subjected to de-unit normalization transformation by standard deviation normalization method. Optionally, the conversion method is usually to subtract the mean value from the data before conversion, and then divide it by the standard deviation to obtain the converted data. The converted data conforms to the standard normal distribution (the mean is 0 and the variance is 1).

It should be noted that those skilled in the art should be able to understand the specific manner of performing the de-unit normalization conversion above, which will not be described in detail here. Further, the manner of performing deunitization and normalization conversion is not limited to only include the above two types.

In the embodiment of the present invention, referring to FIG. 1, in step S120, a similarity analysis is performed on the feature points of the first user and the feature points of the second user among at least two users, and it is judged whether the second user is the first user or not. User's re-entry users, including:

determining similar feature points between the feature points of the first user and the feature points of the second user;

When the ratio between the number of similar feature points and the number of feature points of the first user is greater than a first preset value, it is determined that the second user is a re-entry user of the first user.

Wherein, optionally, determining similar feature points between the feature points of the first user and the feature points of the second user includes:

Selecting the first feature point of the first user;

Calculating a second feature point of the second user with the shortest distance from the first feature point; wherein the target behavior of the behavior feature vector corresponding to the first feature point and the second feature point is the same;

Analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature point are similar feature points.

In the embodiment of the present invention, optionally, the Euclidean algorithm may be selected for the similarity judgment between the first feature point of the first user and the feature point of the second user.

Optionally, the analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature are similar feature points includes:

Acquiring a first weight value when the target behavior corresponding to the first feature point occurs within a preset time period, and acquiring a second weight value when the target behavior corresponding to the second feature point occurs within a preset time period;

determining a weight coefficient according to the first weight value and the second weight value;

calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point;

When it is determined that the similarity value is greater than a second preset value, the first feature point and the second feature are similar feature points.

Further, determining a weight coefficient according to the first weight value and the second weight value includes:

calculating the ratio of the minimum value of the first weight value and the second weight value to the maximum value of the first weight value and the second weight value;

The ratio is determined as the weight coefficient.

Optionally, calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point includes:

The similarity value is calculated according to the following formula:

Si=1-Di/Wi;

Wherein, Si is a similarity value; Di is a distance between the first feature point and the second feature point; Wi is the weight coefficient.

Specifically, by extracting the first feature point of the first user, the second feature point closest to the first feature point is determined in the second user, and the distance is denoted as Di, and according to the target behavior corresponding to the first feature point in the preset The first weight value when it occurs within the time length, and the second weight value when the target behavior corresponding to the second feature point occurs within the preset time length, determine the weight coefficient Wi; according to the calculated distance Di and weight coefficient Wi, That is, the similarity between the first feature point and the second feature point can be calculated.

Optionally, the distance between the first feature point and the second feature point can be calculated according to the above-mentioned determined space-time behavior feature cube, using the coordinate positions of the two feature points in the three dimensions of time, latitude and longitude.

According to the above, through the above method, among the feature points of the second user, the feature point with the closest distance corresponding to each feature point of the first user can be found, and the similarity can be calculated respectively.

Optionally, if there is no feature point of the second user that is closest to the feature point of the first user, the similarity may be marked as 0.

For example, the similarity comparison results of a plurality of feature points of the first user and the second user can be as shown in the following table 3, where B, C and D represent different target behaviors:

table 3

Further, the threshold (second preset value) for similarity judgment can be preset, and when the similarity between the feature points of the second user and the feature points of the first user exceeds the second preset value, it is determined to be similar feature point, otherwise it is not a similar feature point.

In addition, when the first user and the second user carry out re-entry user identification, the threshold (first preset value) reached by the similarity of all feature points can be preset. When the ratio of the number of feature points is greater than the second preset value, it is determined that the first user is highly similar to the second user, and then it is determined that the second user is a re-entry user of the first user.

Using the above similarity judgment principles, the judgment of user behavior similarity has the following three principles:

1. The shorter the distance between two feature points, the higher the similarity;

2. The closer the weight of two feature points is, the higher the similarity is;

3. The more feature points with high similarity between two users, the higher the similarity.

The method for identifying re-entry users described in the embodiments of the present invention utilizes the characteristic that the behaviors of users at a specific time and space have a high degree of similarity, and comprehensively utilizes the three-dimensional data construction of time, space, and user behavior in the data construction stage of the model User feature cube, and use time-dimension data slicing, density clustering algorithm, and behavior intensity threshold determination technology to realize the extraction of typical feature behaviors of users in a specific time space and the construction of behavior feature vectors, and then perform time-dependent analysis of the two user feature vectors. Preprocessing such as domain segmentation and splicing and Euclidean coordinate transformation, and finally, the similarity comparison between two users is carried out through the Euclidean distance algorithm to determine whether the two users are re-entry users.

Using the identification method for re-entry users described in the embodiment of the present invention, the three elements of time, space and behavior are used to perform similarity analysis, and the feature points determined by the behavior feature vector in the spatiotemporal behavior feature cube are used to perform user similarity analysis , compared with a single-dimensional analysis method, it can improve the accuracy of re-entrant user judgment; in addition, by comprehensively modeling data in multiple dimensions, compared with modeling in each dimension separately, it is possible to avoid prone to The problem of misjudgment; further, the simplified user feature vector constructed by various technical means greatly reduces the amount of data for user identity judgment, and improves the efficiency of judging re-entry users.

The embodiment of the present invention also provides an identification device for re-entry users, as shown in Figure 8, including:

A vector acquisition module 810, configured to acquire behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial location, and intensity representation information of the target behavior;

The comparison module 820 is configured to perform similarity analysis on the feature points of the first user and the feature points of the second user among at least two users according to the feature points determined in the spatiotemporal behavior feature cube by the behavior feature vector of each user , judging whether the second user is a re-entry user of the first user;

Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior.

Optionally, the device for identifying re-entrant users, wherein the vector acquisition module 810 includes:

A collection unit 811, configured to collect behavior data of each user; the behavior data includes time, spatial location and intensity representation information of different target behaviors;

A first construction unit 812, configured to construct the spatiotemporal behavior characteristic cube of each user according to the behavior data;

The analysis unit 813 is configured to perform cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determine that the corresponding behavior data whose intensity representation information is greater than a preset intensity threshold is the behavior feature vector;

The second construction unit 814 is configured to delete other feature points in the spatio-temporal behavior feature cube other than the feature points corresponding to the behavior feature vector.

Optionally, in the device for identifying re-entry users, the analysis unit 813 performs cluster analysis on the behavior data in the spatio-temporal behavior characteristic cube, and determines the corresponding behavior whose intensity representation information is greater than a preset intensity threshold The data is the behavior feature vector, including:

Slicing the spatio-temporal behavior feature cube according to the time dimension to form a plurality of slice data;

Clustering the behavioral data in each slice data to determine at least one clustering point;

The intensity representation information of the behavior data corresponding to each cluster point is compared with a preset intensity threshold, and the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is determined as the behavior feature vector.

Optionally, in the device for identifying re-entry users, the intensity indication information is expressed as the cumulative duration of the target behavior within a preset statistical period.

Optionally, the device for identifying a re-entry user, wherein the device further includes:

The conversion module 8101 is used to analyze the behavior feature vector in the spatio-temporal behavior feature cube of the first user before the comparison module 820 performs similarity analysis on the feature points of the first user and the feature points of the second user among the at least two users Carry out de-unit standardization conversion with the behavior feature vector in the spatiotemporal behavior feature cube of the second user, obtain the normalized data vector of the first user and the normalized data vector of the second user; Wherein each normalized data vector corresponds to a feature point;

Wherein, the comparison module 820 performs a similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, specifically:

Perform similarity analysis on the feature points corresponding to the normalized data vector of the first user and the feature points corresponding to the normalized data vector of the second user to determine whether the second user is a re-entry user of the first user.

Optionally, in the device for identifying re-entry users, the comparison module 820 performs a similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, and determines whether the second user A re-entry user who is the first user, specifically:

determining similar feature points between the feature points of the first user and the feature points of the second user;

When the ratio between the number of similar feature points and the number of feature points of the first user is greater than a first preset value, it is determined that the second user is a re-entry user of the first user.

Optionally, the device for identifying re-entry users, wherein the comparison module 820 determines the similar feature points compared with the feature points of the first user and the feature points of the second user, including:

Selecting the first feature point of the first user;

Calculating a second feature point of the second user with the shortest distance from the first feature point; wherein the target behavior of the behavior feature vector corresponding to the first feature point and the second feature point is the same;

Analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature point are similar feature points.

Optionally, in the device for identifying re-entry users, the comparison module 820 analyzes the similarity value between the first feature point and the second feature point, and judges whether the first feature point is similar to the second feature point. Whether the second feature is a similar feature point, including:

Acquiring a first weight value when the target behavior corresponding to the first feature point occurs within a preset time period, and acquiring a second weight value when the target behavior corresponding to the second feature point occurs within a preset time period;

determining a weight coefficient according to the first weight value and the second weight value;

calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point;

When it is determined that the similarity value is greater than a second preset value, the first feature point and the second feature are similar feature points.

Optionally, in the device for identifying re-entrant users, wherein the comparison module 820 determines a weight coefficient according to the first weight value and the second weight value, including:

calculating the ratio of the minimum value of the first weight value and the second weight value to the maximum value of the first weight value and the second weight value;

The ratio is determined as the weight coefficient.

Optionally, in the device for identifying re-entry users, wherein the comparison module 820 calculates a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point, including:

The similarity value is calculated according to the following formula:

Si=1-Di/Wi;

Wherein, Si is a similarity value; Di is a distance between the first feature point and the second feature point; Wi is the weight coefficient.

Optionally, the device for identifying re-entry users, wherein, before de-uniting and normalizing the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user , the conversion module 8101 is also used for:

In the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the time distribution dimension of the behavioral feature vector of the first user and the behavioral feature vector of the second user is determined to determine the time domain segmentation point;

Segment and assemble the spatiotemporal behavior characteristic cube of the first user and the spatiotemporal behavior characteristic cube of the second user, so that the spatiotemporal behavior characteristic cube of the first user and the second user's spatiotemporal behavior characteristic cube after segmentation and assembling The spatio-temporal behavior characteristic cube uses the time corresponding to the time domain segmentation point as the starting time point;

Wherein, the conversion module 8101 performs de-unit normalization conversion on the behavior feature vector in the spatiotemporal behavior feature cube of the first user and the behavior feature vector in the spatiotemporal behavior feature cube of the second user, including:

De-unit normalization is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube after segmentation and assembly.

Optionally, in the device for identifying re-entry users, the transformation module 8101 constructs the spatio-temporal behavior feature cube according to the behavior feature vector of the first user and the behavior feature vector of the second user, the behavior feature of the first user Vector and the time distribution dimension of the second user's behavior feature vector, determine the time domain segmentation point, including:

According to the time dimension, the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vectors in the spatio-temporal behavior feature cube of the second user respectively accumulate the intensity representation information corresponding to the same target behavior;

Draw a first intensity change curve of the behavior feature vector of the first user and draw a second intensity change curve of the behavior feature vector of the second user according to the accumulated maximum intensity information value corresponding to each target behavior;

The lowest point of the first intensity variation curve and the second intensity variation curve is selected as the time domain segmentation point.

Optionally, in the device for identifying re-entry users, the transformation module 8101 deunits the behavior feature vectors in the spatiotemporal behavior feature cube of the first user and the behavior feature vector in the spatiotemporal behavior feature cube of the second user Normalized transformations, including:

By means of a dispersion normalization method or a standard deviation normalization method, de-unit normalization conversion is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube respectively.

An embodiment of the present invention also provides an identification device, as shown in FIG. 9 , including: a processor 901; and a memory 903 connected to the processor 901 through a bus interface 902, and the memory 903 is used to store the processing The processor 901 calls and executes the programs and data stored in the memory 903 .

Wherein, the transceiver 904 is connected to the bus interface 902, and is used to receive and send data under the control of the processor 901. Specifically, the processor 901 is used to read the program in the memory 903, and execute the following process:

Obtaining behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior;

According to the feature points determined by the behavior feature vector of each user in the spatio-temporal behavior feature cube, the feature points of the first user and the feature points of the second user among at least two users are analyzed for similarity, and the second user is judged. Whether the user is a re-entry user of the first user;

Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior.

Optionally, the identification device, wherein the processor 901 acquires behavior feature vectors of at least two users, includes:

Collect behavior data of each user; the behavior data includes time, space position and intensity representation information of different target behaviors;

Constructing the spatiotemporal behavior characteristic cube of each user according to the behavior data;

performing cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determining that the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is the behavior feature vector;

Deleting other feature points in the spatio-temporal behavior feature cube other than the feature points corresponding to the behavior feature vector.

Optionally, in the identification device, the processor 901 performs cluster analysis on the behavior data in the spatio-temporal behavior characteristic cube, and determines that the corresponding behavior data whose intensity representation information is greater than a preset intensity threshold is the Behavioral feature vectors, including:

Slicing the spatio-temporal behavior feature cube according to the time dimension to form a plurality of slice data;

Clustering the behavioral data in each slice data to determine at least one clustering point;

The intensity representation information of the behavior data corresponding to each cluster point is compared with a preset intensity threshold, and the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is determined as the behavior feature vector.

Optionally, in the identification device, the intensity indication information is expressed as the cumulative duration of the target behavior within a preset statistical period.

Optionally, in the identification device, the processor 901 is further configured to:

The behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user are deunited and standardized to obtain the normalized data vector of the first user and the normalized data vector of the second user ; Each normalized data vector corresponds to a feature point;

Wherein, the processor 901 performs a similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, including:

Perform similarity analysis on the feature points corresponding to the normalized data vector of the first user and the feature points corresponding to the normalized data vector of the second user to determine whether the second user is a re-entry user of the first user.

Optionally, in the identification device, the processor 901 performs a similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, and determines whether the second user is the second user. A user's re-entry user, including:

determining similar feature points between the feature points of the first user and the feature points of the second user;

When the ratio between the number of similar feature points and the number of feature points of the first user is greater than a first preset value, it is determined that the second user is a re-entry user of the first user.

Optionally, in the identification device, wherein the processor 901 determines similar feature points between the feature points of the first user and the feature points of the second user, including:

Selecting the first feature point of the first user;

Calculating a second feature point of the second user with the shortest distance from the first feature point; wherein the target behavior of the behavior feature vector corresponding to the first feature point and the second feature point is the same;

Analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature point are similar feature points.

Optionally, in the identification device, the processor 901 analyzes the similarity value between the first feature point and the second feature point to determine whether the first feature point and the second feature are similar feature points, including:

Acquiring a first weight value when the target behavior corresponding to the first feature point occurs within a preset time period, and acquiring a second weight value when the target behavior corresponding to the second feature point occurs within a preset time period;

determining a weight coefficient according to the first weight value and the second weight value;

calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point;

When it is determined that the similarity value is greater than a second preset value, the first feature point and the second feature are similar feature points.

Optionally, in the identification device, wherein the processor 901 determines a weight coefficient according to the first weight value and the second weight value, including:

calculating the ratio of the minimum value of the first weight value and the second weight value to the maximum value of the first weight value and the second weight value;

The ratio is determined as the weight coefficient.

Optionally, in the identification device, wherein the processor 901 calculates the similarity value according to the weight coefficient and the distance between the first feature point and the second feature point, including:

The similarity value is calculated according to the following formula:

Si=1-Di/Wi;

Wherein, Si is a similarity value; Di is a distance between the first feature point and the second feature point; Wi is the weight coefficient.

Optionally, in the identification device, before the processor 901 performs deunit normalization transformation on the behavior feature vectors in the first user’s spatiotemporal behavior feature cube and the behavior feature vector in the second user’s spatiotemporal behavior feature cube, Also used for:

In the spatio-temporal behavioral feature cube constructed according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the time distribution dimension of the behavioral feature vector of the first user and the behavioral feature vector of the second user is determined to determine the time domain segmentation point;

Segment and assemble the spatiotemporal behavior characteristic cube of the first user and the spatiotemporal behavior characteristic cube of the second user, so that the spatiotemporal behavior characteristic cube of the first user and the second user's spatiotemporal behavior characteristic cube after segmentation and assembling The spatio-temporal behavior characteristic cube uses the time corresponding to the time domain segmentation point as the starting time point;

Wherein, the processor 901 performs deunit normalization transformation on the behavior feature vector in the spatiotemporal behavior feature cube of the first user and the behavior feature vector in the spatiotemporal behavior feature cube of the second user, including:

De-unit normalization is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube after segmentation and assembly.

Optionally, in the recognition device, the processor 901 constructs the spatio-temporal behavior feature cube according to the behavior feature vector of the first user and the behavior feature vector of the second user, the behavior feature vector of the first user and the behavior feature vector of the second user The time distribution dimension of the user's behavior feature vector determines the time domain segmentation point, including:

According to the time dimension, the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vectors in the spatio-temporal behavior feature cube of the second user respectively accumulate the intensity representation information corresponding to the same target behavior;

Draw a first intensity change curve of the behavior feature vector of the first user and draw a second intensity change curve of the behavior feature vector of the second user according to the accumulated maximum intensity information value corresponding to each target behavior;

The lowest point of the first intensity variation curve and the second intensity variation curve is selected as the time domain segmentation point.

Optionally, in the identification device, wherein, the processor 901 performs deunit normalization transformation on the behavior feature vector in the first user’s spatiotemporal behavior feature cube and the behavior feature vector in the second user’s spatiotemporal behavior feature cube, including :

By means of a dispersion normalization method or a standard deviation normalization method, de-unit normalization conversion is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube respectively.

Wherein, in FIG. 9 , the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by the processor 901 and various circuits of the memory represented by the memory 903 are linked together. The bus architecture can also link together various other circuits such as peripherals, voltage regulators, and power management circuits, etc., which are well known in the art and therefore will not be further described herein. The bus interface provides the interface. Transceiver 904 may be a plurality of elements, including a transmitter and a receiver, providing a means for communicating with various other devices over transmission media. The processor 901 is responsible for managing the bus architecture and general processing, and the memory 903 can store data used by the processor 901 when performing operations.

Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiments can be implemented by hardware, or can be completed by instructing the relevant hardware through a program, and the program includes instructions for executing some or all of the steps of the above-mentioned method; and The program can be stored in a readable storage medium, and the storage medium can be any form of storage medium.

In addition, a specific embodiment of the present invention also provides a readable storage medium on which a computer program is stored, wherein, when the program is executed by a processor, the steps of the method for identifying re-entry users as described in any one of the above items are implemented.

Specifically, the readable storage medium is applied to the above-mentioned identification device. When applied to the identification device, the execution steps in the method for identifying re-entry users are as described in detail above, and will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed methods and devices may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, each unit may be physically included separately, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or in the form of hardware plus software functional units.

The above-mentioned integrated units implemented in the form of software functional units may be stored in a computer-readable storage medium. The above-mentioned software functional units are stored in a storage medium, and include several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) execute some steps of the sending and receiving methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM for short), random access memory (Random Access Memory, RAM for short), magnetic disk or optical disk, etc., which can store program codes. medium.

What has been described above is a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, some improvements and modifications can also be made without departing from the principle of the present invention. These improvements and modifications should also be considered as Be the protection scope of the present invention.

Claims (16)

1. An identification method for re-entry users, characterized in that, comprising: Obtaining behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior; According to the feature points determined by the behavior feature vector of each user in the spatio-temporal behavior feature cube, the feature points of the first user and the feature points of the second user among at least two users are analyzed for similarity, and the second user is judged. Whether the user is a re-entry user of the first user; Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior. 2. the identification method of re-entrant network user according to claim 1, is characterized in that, the behavior characteristic vector of described acquisition at least two users, comprises: Collect behavior data of each user; the behavior data includes time, space position and intensity representation information of different target behaviors; Constructing the spatiotemporal behavior characteristic cube of each user according to the behavior data; performing cluster analysis on the behavior data in the spatio-temporal behavior feature cube, and determining that the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is the behavior feature vector; Deleting other feature points in the spatio-temporal behavior feature cube other than the feature points corresponding to the behavior feature vector. 3. The method for identifying re-entry users according to claim 2, wherein the behavior data in the spatio-temporal behavior characteristic cube is clustered and analyzed, and the corresponding intensity representation information is determined to be greater than the preset intensity threshold. Behavior data is the behavior feature vector, including: Slicing the spatio-temporal behavior feature cube according to the time dimension to form a plurality of slice data; Clustering the behavioral data in each slice data to determine at least one clustering point; The intensity representation information of the behavior data corresponding to each cluster point is compared with a preset intensity threshold, and the corresponding behavior data whose intensity representation information is greater than the preset intensity threshold is determined as the behavior feature vector. 4. The method for identifying re-entry users according to any one of claims 1 to 3, wherein the intensity indication information is expressed as the cumulative duration of the target behavior within a preset statistical period. 5. the identification method of re-entrant network user according to claim 1, is characterized in that, before carrying out the similarity analysis to the feature point of the first user and the feature point of the second user in at least two users, described method also include: The behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user are deunited and standardized to obtain the normalized data vector of the first user and the normalized data vector of the second user ; Each normalized data vector corresponds to a feature point; Wherein, the similarity analysis is performed on the feature points of the first user and the feature points of the second user among at least two users, including: Perform similarity analysis on the feature points corresponding to the normalized data vector of the first user and the feature points corresponding to the normalized data vector of the second user to determine whether the second user is a re-entry user of the first user. 6. The method for identifying re-entry users according to claim 1, characterized in that, performing similarity analysis on the feature points of the first user and the feature points of the second user among at least two users, and judging that the second user Whether it is a re-entry user of the first user, including: determining similar feature points between the feature points of the first user and the feature points of the second user; When the ratio between the number of similar feature points and the number of feature points of the first user is greater than a first preset value, it is determined that the second user is a re-entry user of the first user. 7. The method for identifying re-entry users according to claim 6, wherein said determining the similar feature points of the feature points of the first user compared with the feature points of the second user comprises: Selecting the first feature point of the first user; Calculating a second feature point of the second user with the shortest distance from the first feature point; wherein the target behavior of the behavior feature vector corresponding to the first feature point and the second feature point is the same; Analyzing the similarity value between the first feature point and the second feature point, and judging whether the first feature point and the second feature point are similar feature points. 8. The identification method of re-entrant network users according to claim 7, characterized in that, said analyzing the similarity value between said first feature point and said second feature point, judging said first feature point and said second feature point Whether the second feature is a similar feature point, including: Acquiring a first weight value when the target behavior corresponding to the first feature point occurs within a preset time period, and acquiring a second weight value when the target behavior corresponding to the second feature point occurs within a preset time period; determining a weight coefficient according to the first weight value and the second weight value; calculating a similarity value according to the weight coefficient and the distance between the first feature point and the second feature point; When it is determined that the similarity value is greater than a second preset value, the first feature point and the second feature are similar feature points. 9. The method for identifying re-entry users according to claim 8, wherein, according to the first weight value and the second weight value, determining a weight coefficient includes: calculating the ratio of the minimum value of the first weight value and the second weight value to the maximum value of the first weight value and the second weight value; The ratio is determined as the weight coefficient. 10. The identification method of re-entrant network user according to claim 8, is characterized in that, according to the distance between described weight coefficient and described first feature point and described second feature point, calculate similarity value, comprise : The similarity value is calculated according to the following formula: Si=1-Di/Wi; Wherein, Si is a similarity value; Di is a distance between the first feature point and the second feature point; Wi is the weight coefficient. 11. the identification method of re-entrant network user according to claim 5 is characterized in that, the behavior feature vector in the spatiotemporal behavior characteristic cube of the first user and the behavior characteristic vector in the spatiotemporal behavior characteristic cube of the second user are removed Before unit normalization conversion, the method also includes: In the spatio-temporal behavior feature cube constructed according to the behavior feature vector of the first user and the behavior feature vector of the second user, the time distribution dimension of the behavior feature vector of the first user and the behavior feature vector of the second user is determined to determine the time domain segmentation point; Segment and assemble the spatiotemporal behavior characteristic cube of the first user and the spatiotemporal behavior characteristic cube of the second user, so that the spatiotemporal behavior characteristic cube of the first user and the second user's spatiotemporal behavior characteristic cube after segmentation and assembling The spatio-temporal behavior characteristic cube uses the time corresponding to the time-domain segmentation point as the starting time point; Wherein, the de-unit standardization transformation is performed on the behavior feature vector in the spatio-temporal behavior feature cube of the first user and the behavior feature vector in the spatio-temporal behavior feature cube of the second user, including: Perform deunit normalization transformation on the behavior feature vectors in the first user's spatiotemporal behavior feature cube and the second user's spatiotemporal behavior feature cube after segmentation and assembly. 12. The method for identifying re-entry users according to claim 11, characterized in that, in the spatio-temporal behavioral feature cube constructed respectively according to the behavioral feature vector of the first user and the behavioral feature vector of the second user, the behavior of the first user The feature vector and the time distribution dimension of the second user's behavior feature vector determine the time domain segmentation point, including: According to the time dimension, the behavior feature vectors in the spatio-temporal behavior feature cube of the first user and the behavior feature vectors in the spatio-temporal behavior feature cube of the second user respectively accumulate the intensity representation information corresponding to the same target behavior; Draw a first intensity change curve of the behavior feature vector of the first user and draw a second intensity change curve of the behavior feature vector of the second user according to the accumulated maximum intensity information value corresponding to each target behavior; The lowest point of the first intensity change curve and the second intensity change curve is selected as the time domain segmentation point. 13. the identification method of re-entrant network user according to claim 5 is characterized in that, the behavior feature vector in the spatiotemporal behavior characteristic cube of the first user and the behavior characteristic vector in the spatiotemporal behavior characteristic cube of the second user are removed Unit normalization conversions, including: By means of a dispersion normalization method or a standard deviation normalization method, de-unit normalization conversion is performed on the behavior feature vectors in the first user's spatio-temporal behavior feature cube and the second user's spatio-temporal behavior feature cube respectively. 14. An identification device for re-entry users, characterized in that it comprises: A vector acquisition module, configured to acquire behavioral feature vectors of at least two users; the behavioral feature vectors record the occurrence time, spatial position and intensity representation information of the target behavior; The comparison module is used to perform similarity analysis on the feature points of the first user and the feature points of the second user among at least two users according to the feature points determined in the spatio-temporal behavior feature cube by the behavior feature vector of each user, judging whether the second user is a re-entry user of the first user; Wherein the spatio-temporal behavior feature cube takes time, the longitude of the spatial position, and the latitude of the spatial position as coordinates, and the feature points corresponding to the behavior feature vector are distributed in the spatio-temporal behavior feature cube according to the time and space position of the target behavior. 15. An identification device, characterized in that it comprises: a processor, a memory, and a program stored on the memory and operable on the processor, when the program is executed by the processor, it realizes The method for identifying re-entry users described in any one of 1 to 13. 16. A readable storage medium, characterized in that a program is stored on the readable storage medium, and when the program is executed by a processor, the identification of re-entrant network users according to any one of claims 1 to 13 is realized steps in the method.

Images