CN102135983A - Group dividing method and device based on network user behavior - Google Patents

Abstract

The invention discloses a group division method and device based on network user behavior, wherein the method includes: determining a behavior characteristic parameter set according to user requirements; constructing a user type space according to the determined behavior characteristic parameter set; The data is read in the user behavior server, and after normalizing various parameters of each user, a user constellation diagram is formed in the user type space, and each point in the constellation diagram is initialized and the detection range is determined. The moving distance vector is calculated based on the distance between the point and itself, and the detection range is updated according to the moving distance vector, and the above steps are repeated until all points do not move; according to the initial detection range, the adjacent points form a user group.

Description

Group division method and device based on network user behavior

technical field

The invention relates to the field of network user behavior analysis, in particular to a method and device for analyzing user business behavior characteristics and rules.

Background technique

At present, with the rapid increase of Internet users, Internet services are also enriched, and different services correspond to different groups of people. Therefore, to make a service more suitable for its corresponding users, it is necessary to divide the users who use the service into groups, and then improve the existing services in a targeted manner according to different user groups. User group segmentation technology is a key technology for future user behavior analysis.

The user group division method in the prior art is mainly based on the division method based on logical links based on LM (Link Mining), which is not conducive to the cross-use of multiple services in the telecommunications network, and the situation of multiple users of the Internet and non-Internet At the same time, the excessive program complexity also limits the scope of application of this method.

Contents of the invention

The embodiment of the present invention provides a group division method and device based on network user behavior, which makes the user group division more in line with the monitoring situation of the telecommunication network and meets the special requirements of users, and at the same time expands the scope of use.

The embodiment of the present invention discloses a group division method based on network user behavior, including: determining a behavior characteristic parameter set; constructing a user type space according to the determined behavior characteristic parameter set, reading data from a user behavior server, and After the various parameters of each user are normalized, a user constellation diagram is formed in the user type space, each point in the constellation diagram is initialized and the detection range is determined, and the moving distance vector is calculated according to the distance between the points within the detection range of the point and itself , and at the same time update the detection range according to the moving distance vector, repeat the above steps until all points do not move; according to the initial detection range, the adjacent points form a user group.

The embodiment of the present invention also discloses a device for group division based on network user behavior, which is characterized in that it includes:

A parameter set determining unit, configured to determine a behavior characteristic parameter set;

a space determining unit, configured to determine the user type space according to the behavior characteristic parameter set;

The clustering unit reads data from the user behavior server, and after normalizing the various parameter values of each user, forms a user constellation diagram in the user type space and determines the detection range. Calculate the moving distance vector based on its own distance, and at the same time update the detection range according to the moving distance vector, repeat the above steps until all points do not move

The group division unit is configured to form a user group from adjacent points according to the initial detection range.

It can be seen from the above-mentioned embodiments of the present invention that the user behavior space in the embodiments of the present invention can be changed to meet the requirements for cross-use of multiple services in the telecommunication network, and user requirements are introduced as the corresponding relationship of behavior characteristic parameter sets, indicating the importance of parameters. The degree of flexibility, therefore taking into account the special requirements of users. At the same time, the clustering method of the present invention is not limited to the group division of network users, which expands the scope of application.

Description of drawings

Fig. 1 is a flow chart of an embodiment of the method for realizing user group division in the present invention;

FIG. 2 is a flow chart of another embodiment of the method for realizing network selection in the present invention;

Fig. 3 is the flowchart of distance vector clustering method of the present invention;

Fig. 4 is a structural diagram of an embodiment of the device for realizing user group division according to the present invention.

Detailed ways

Embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Referring to Fig. 1, it is a flowchart of an embodiment of a method for network selection in the present invention, including the following steps:

Step 101: determining a behavior characteristic parameter set;

Step 102: Construct a user type space according to the determined behavior characteristic parameter set;

Step 103: Read data from the user behavior server, normalize various parameters of each user, form a user constellation diagram in the user type space, initialize each point in the constellation diagram and determine the detection range, according to the The distance between the points within the point detection range and itself is calculated to calculate the moving distance vector, and at the same time, the detection range is updated according to the moving distance vector, and the above steps are repeated until all points do not move;

Step 104: according to the initial detection range, form a user group with adjacent points;

In the above embodiment, data is read from the user behavior server in user group division, and after normalizing various parameter values of each user, a user constellation diagram is formed in the user type space and the detection range is determined. According to the detection range at this point Calculate the moving distance vector based on the distance between the points within and itself, and at the same time update the detection range according to the moving distance vector, repeat the above steps until all points do not move, so as to achieve the user clustering effect. Among them, the user behavior space in the embodiment of the present invention can be changed to meet the requirements of limited data, and user requirements are introduced as the corresponding relationship of behavior characteristic parameter sets to indicate the importance of parameters, so the requirements of business developers are considered. At the same time, the clustering method of the present invention is not limited to the group division of network users, which expands the scope of application.

Please refer to FIG. 2 , which is a flow chart of another embodiment of the method for realizing user group division in the present invention. In this embodiment, the information concerned by the user is obtained through the user interface. In this embodiment, behavior characteristic parameters are stored in UBAS, such as user static type and user dynamic topic, and IP addresses are also divided into regions. When users need to divide user groups, the following steps are included:

Step 201: The user group division device UGDD (User Group Diving Device) obtains the user requirements and each behavior characteristic parameter from the current UBAS (User Behavior analysis Server), and determines according to the corresponding relationship between the user requirement and the behavior characteristic parameter set The corresponding set of behavioral feature parameters.

Among them, UGDD is a component of the current UBAS, and each UBAS has at least one UGDD.

In order to determine the corresponding relationship between the behavioral feature parameter set and the user’s requirements, the user’s hot topics and attention levels are read in advance before the user group is divided: "Diaoyu Island, 1.0", "Japan, 0.6", "China, 0.8" user static Type and attention: "blog, 0.5", "news, 0.9", region: "Henan", time: 20100712 as user request data and various behavior characteristic parameters that have been processed in UBAS.

Therefore, the behavior characteristic parameter set selected in this embodiment is:

Table 1 Behavioral feature parameter set obtained according to user requirements

Step 202: UGDD, according to the behavior characteristic parameter set, constructs user type Euclidean space, and determines the normalization standard at the same time.

The user type Euclidean space constructed above is in the behavior characteristic parameter set, except for the two parameters of "time" and "region", there are still 5 parameters, so at this time, a 5-dimensional coordinate Euclidean space is constructed, and the 5 coordinate axes are: Diaoyu Island axis , Japanese axis, Chinese axis, blog axis and news axis.

The normalization standard S=(1.0, 0.6, 0.8, 0.5, 0.9) is determined above.

Step 203: UGDD reads data from UBAS, normalizes user data by normalization standard, forms constellation diagram in user type space, uses distance vector clustering algorithm to cluster, and gathers similar users

The above UGDD reads data from UBAS, and normalizes the user data through the normalization standard as follows:

据归一化标准S＝(1.0，0.6，0.8，0.5，0.9)，由公式

将这T个用户均归一化，得到用户i为IP_i＝(p_i1，p_i2，...，p_in)，这时将这T个用户点入到用户类型空间中形成星座图，每一个点根据公式，计算其初始探查范围w₀，令探查范围w_i＝w₀，最小探查范围w_min＝w₀/4。The result of UGDD reading the data of 20100712 Henan in UBAS is that there are a total of T users, and each of the T users has its own value for the five coordinate axes in 202. At this time, user i can be expressed as

According to the normalization standard S=(1.0,0.6,0.8,0.5,0.9), by the formula

The T users are all normalized, and the user i is obtained as IP _i =(p _i1 , p _i2 ,..., _pin ), and these T users are entered into the user type space to form a constellation diagram, Each point according to the formula, Calculate the initial detection range w ₀ , set the detection range w _i =w ₀ , and the minimum detection range w _min =w ₀ /4.

For the above-mentioned distance vector clustering algorithm clustering, please refer to Figure 3 for the specific implementation of clustering similar users, including the following steps:

Step 301: Determine the n-dimensional Euclidean space, L represents the length of the vector, d _ij represents the distance between point i and j;

In this embodiment, the n-dimensional Euclidean space is the user type space

Step 302: Calculate the moving path;

If the distance d _ij <w between point i and point j, then point j is called the adjacent point of point i or point i is the adjacent point of j, and the moving path d i of point _i is passed by the formula

calculated;

Step 303: clustering movement;

Let L _di be the length of the moving path d _i , let ID _i move as ID _i +d _i , and update the detection range at the same time, so that w _i =L _di ; carry out the above-mentioned move once for each point in the space;

Step 304: move end judgment;

When all points do not move, the clustering ends, and similar users have been concentrated after the end.

Step 203: UGDD forms a user group with adjacent points according to the initial detection range, and feeds back the result to UBAS at the same time;

The initial detection range is w ₀ , each point i searches for neighboring points according to the initial detection range, sets point i and its neighboring points as a group, repeats the above detection process, if the point has been classified into a group, no detection is performed, After all the nodes are classified into the group, it ends; the divided network user groups are obtained.

UGDD stores each divided user group into UBAS.

It can be seen from the above embodiments that the user behavior space of the present invention can be changed. When the behavior characteristic parameter data in UBAS is less, the generated space will also be smaller, which meets the requirements of limited data and introduces user requirements as behavior characteristics. The correspondence between parameter sets indicates the degree of importance between parameters, so the requirements of business developers are taken into account. At the same time, the clustering method of the present invention is not limited to the group division of network users, which expands the scope of application.

The above embodiments use the current user behavior analysis server UBAS as the source of behavior feature data, but the present invention is not limited to these two analysis models. For example, when the behavior feature is not a network behavior feature but other behavior features, its user group division The method is similar to the above embodiment, and will not be repeated here. Please refer to FIG. 4 , which is a structural diagram of an embodiment of a device for realizing user group division in the present invention, the device includes a parameter set determination unit 401, a space determination unit 402, a clustering unit 403, and a group division unit 404

A parameter set determining unit 401, configured to determine a behavior characteristic parameter set according to user requirements;

A space determining unit 402, configured to determine the user type space according to the behavior characteristic parameter set determined by the parameter set determining unit 401;

The clustering unit 403, according to the space determination unit 402, from the parameter set, after normalizing various parameter values of each user, forms a user constellation diagram in the user type space and determines the detection range, according to the detection range at this point The distance between the point and itself is used to calculate the moving distance vector, and at the same time, the detection range is updated according to the moving distance vector, and the above steps are repeated until all points do not move

The unit also includes:

The group division unit 404 is configured to form a user group from adjacent points by using the initial detection range according to the result obtained by the clustering unit 403 .

In this embodiment, the user behavior space is not limited to the requirements of Internet behavior characteristics, and user requirements are introduced as the corresponding relationship of behavior characteristic parameter sets to indicate the importance of parameters, which meets the special requirements of users and takes into account the telecommunications The case of cross-use of multiple services on the network. At the same time, the clustering method of the present invention is not limited to the group division of network users, which expands the scope of application.

Those of ordinary skill in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium, and the program includes the following Steps: Determine the behavioral characteristic parameter set according to the user's requirements; construct the user type space according to the determined behavioral characteristic parameter set; Form the user constellation diagram in the type space, initialize each point in the constellation diagram and determine the detection range, calculate the moving distance vector according to the distance between the points within the detection range of the point and itself, and update the detection range according to the moving distance vector, repeat the above Steps until all points do not move; according to the initial exploration range, the adjacent points form a user group. The storage medium, such as: ROM/RAM, magnetic disk, optical disk, etc.

The above is only a preferred embodiment of the present invention, it should be pointed out that, for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications can also be made. It should be regarded as the protection scope of the present invention.

Claims (9)

1. colony's division methods of a user behavior Network Based is characterized in that, comprises

Determine the cybernetics control number collection;

According to fixed cybernetics control number collection, make up the user type space;

From the user behavior server, read in data, after the multiple parameter normalization of each user, in the user type space, form user's planisphere, each point in the planisphere is carried out initialization and determines the scope of detecting, according to detect point in the scope and the distance calculation displacement vector of itself at this point, upgrade simultaneously and detect scope, repeat above step, have a few mobile up to institute according to the displacement vector;

According to initially detecting scope, contiguous point is constituted a customer group.

2. method according to claim 1 is characterized in that, also comprises before described definite cybernetics control number:

In predetermined cybernetics control number scope, according to user's the needs setting and the corresponding relation of current behavior characteristic parameter collection;

Then described definite current behavior characteristic parameter collection is for obtaining the dynamic topic of user, user's static types and IP address from user behavior analysis server UBAS (User Behavior Analysis Server), according to user's the needs and the corresponding relation of current behavior characteristic parameter collection, determine the corresponding cybernetics control number collection of needs with the user;

3. method according to claim 2 is characterized in that, described cybernetics control number scope comprises:

The dynamic topic of user is by being used the determined keyword of Network by the user in a period of time in the user behavior server, this keyword can since the user who is analyzed for geographic position and time of use business different and different;

User's static types is meant the professional type of using, and comprising: news, physical culture, amusement, science and technology, finance and economics, automobile, number, mobile phone, house property, recreation, tourism, education, 13 classes of blog; These classifications are to add up by the historical data of user behavior server to draw.

4. method according to claim 2 is characterized in that, the user requires to comprise:

To degree of concern some or the dynamic topic of several users or user's static types, the selection of time period, the selection of geographic area scope.

5. method according to claim 1 is characterized in that, described definite cybernetics control number collection comprises:

Obtain with the user from UBAS and to require corresponding behavioural analysis parameter set.

6. according to claim 1,2 or 5 any described methods, it is characterized in that,, make up the user type space and comprise according to fixed cybernetics control number collection:

This user type space is a theorem in Euclid space;

With the orthogonal axis in the dynamic topic structure of user's static types and user in the parameter set space;

User's static types and the dynamic topic quantity of user sum n are called range parameter;

With the attention rate of each user's static types and the dynamic topic of user in the parameter set as data normalization standard S=(s ₁, s ₂..., s _n);

7. according to claim 1,2 or 5 any described methods, it is characterized in that, describedly from the user behavior server, read in data, after the multiple parameter values normalization of each user, form user's planisphere and determine the scope of detecting in the user type space, according to the point and the distance calculation displacement vector of itself detected at this point in the scope, scope is detected in renewal according to the displacement vector simultaneously, repeat above step, up to have a few not move and comprise:

To from the user behavior server, reading in data, after the multiple parameter values normalization of each user, in the user type space, form user's planisphere and determine the scope of detecting according to:

From the user behavior server, read each user i data and indicate this user, obtain data set with the IP address

According to number

According to normalization standard S=(s ₁, s ₂..., s _n), by formula

Normalization,

Each parameter of user i obtains IP _i=(p _I1, p _I2..., p _In) simultaneously, click and enter in the user type space,

Form planisphere, each is put according to formula,

Calculate its initial spy

Look into scope w ₀, scope w is detected in order _i=w ₀, minimum is detected scope w _Min=w ₀/ 4.

To according to point and the distance calculation displacement vector of itself detected at this point in the scope, upgrade according to the displacement vector simultaneously and detect scope, repeat above step, up to have a few mobilely according to the distance vector clustering algorithm, carry out mobile cluster;

The vector clustering algorithm of adjusting the distance carries out according to following steps:

A) in this n dimension theorem in Euclid space, L represents vector length, d _IjThe distance of expression point i and j;

B) if the some i with the some j apart from d _Ij＜w claims that then neighbor point or the some i of some j for some i is j

Neighbor point, the some i mobile route d _iBy

Calculate and get, simultaneously, make L _DiBe mobile route d _iLength, make ID _iBe ID _i+ d _iMove, upgrade simultaneously and detect scope, make w _i=L _DiEach point in the space is carried out once above-mentioned moving;

C) repeat b), when institute has a few when not mobile the cluster end;

8. according to claim 1,2,5,6 or 7 any described methods, it is characterized in that described basis is initially detected scope, contiguous point constituted a customer group comprise:

Initially the scope of detecting is w ₀, each some i searches neighbor point according to initially detecting scope, will put i and neighbor point thereof and be made as a group, repeats above-mentioned probe procedure, if this point has been included into a group and has not then detected, all nodes all are included into the group back to be finished; The network user group who obtains dividing.

9. the colony's division methods and the device of a user behavior Network Based is characterized in that, comprising:

The parameter set determining unit is used for determining the cybernetics control number collection;

The space determining unit is used for determining the user type space according to the cybernetics control number collection;

Cluster cell, from the user behavior server, read in data, after the multiple parameter values normalization of each user, in the user type space, form user's planisphere and determine the scope of detecting, according to detect point in the scope and the distance calculation displacement vector of itself at this point, upgrade simultaneously and detect scope, repeat above step, have a few mobile up to institute according to the displacement vector

Group's division unit is used for according to initially detecting scope contiguous point being constituted a customer group.

Images