CN105678457A - Method for evaluating user behavior on the basis of position mining - Google Patents

Abstract

The invention discloses a method for evaluating user behavior on the basis of position mining. The method comprises following steps: a. obtaining position information to obtain tracking data and performing pretreatment to the tracking data; b. obtaining time distribution of a user at different places in different time through the tracking data to obtain a behavior pattern of the user; c. updating a user preference model by use of historical records of the user behavior pattern and determining whether a user behavior is an abnormal behavior according to the user preference model and calculating user behavior scores. By use of the method, the user behavior pattern is obtained from user original messy tracking data, daily behavior rules of the user can be revealed, and the possibility of cheat in traditional attendance checking is greatly reduced.

Description

Based on the Trustworthy user behaviour method that place is excavated

Technical field

The present invention relates to a kind of Trustworthy user behaviour method excavated based on place, belong to Data Mining.

Background technology

Along with the development of mobile Internet, mobile phone etc. moves equipment and popularizes gradually. The mobile equipment such as current mobile phone is provided with GPS, or has network positions function, facilitates user automatically to record the stroke of every day. User's different time sections every day residence time destribution in different location, reflects the Behavior law of this user.

Based on user trajectory data data mining technology constantly development in, it is typically employed in the fields such as popular place recommendation, the application in enterprise field is also fewer, for instance, 2013 " Kunming University of Science and Technology " disclosed user based on mobile phone location data goes on a journey law-analysing.

By digging user track data, enterprise can assess the work behavior of employee easily. This technology is not solely restricted to business administration, it is also possible to being applied in any needs track data to do the Trustworthy user behaviour field supported. But the examination employee's work performance outside of the existing traditional forms of enterprises is checked card generally by work attendance and realized, and this method needs manual maintenance record, the schemes such as fingerprint machine are adopted to also need to put into additional hardware cost, it is possible to the place of identification is also restrained.

Summary of the invention

It is an object of the invention to the problems referred to above overcoming prior art to exist, it is provided that a kind of Trustworthy user behaviour method excavated based on place. The present invention from the original mixed and disorderly track data of user, can obtain the behavioral pattern of user, and accurate response goes out the Behavior law of user's every day, greatly reduces traditional work attendance and checks card the possibility of in violation of rules and regulations cheating.

For achieving the above object, the technical solution used in the present invention is as follows:

A kind of Trustworthy user behaviour method excavated based on place, it is characterised in that: utilize the behavioral pattern of trajectory data mining user, analyze the Behavior law of user, user behavior is estimated.

Described method specifically includes following steps:

A, acquisition positional information obtain track data, and to track data prediction;

B, by track data obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user;

C, utilize the historical record of user behavior pattern to update user preferences modeling, determine whether Deviant Behavior according to user preferences modeling, calculate user behavior score.

In described step a, mobile equipment reports track data to be L={l₁,l₂,…,l_n, wherein l_i=(lat_i,long_i,time_i) represent longitude and latitude and time, first remove and track data repeats a little, again through Kalman filter rate of filtration abnormity point, smooth track data, make track data closer to real travel path.

In described step b, adopt based on seasonal effect in time series clustering algorithm, obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user.

Described step b specifically includes:

B1, choose two parameter value place maximum magnitude D_maxWith effective place time span T;

B2, successively track data being carried out following process: when the distance of adjacent two tracing points is less than place range threshold D, two tracing points merge into a new tracing point, new tracing point participates in processing next time; When the distance of adjacent two tracing points is more than place range threshold D_maxTime, and the time span of previous tracing point more than effective place time threshold T time, this tracing point is an effective place, represent p={lat, lng, start_ts, end_ts};

B3, user count out as M effectively, within one day, are divided into N number of time period, and user behavior pattern is expressed as M × N matrix P=[f_ij], every a line represents certain effective place of user, and certain time period in one day, f are shown in each list_ijExpression rests on the probability in i-th place in the jth time period.

In described step c, user preferences modeling is drawn by following method:

C1, by Cosine coefficient, define user behavior pattern distance function. For behavioral pattern X and behavioral pattern Y, the place distribution vector of i-th time period is X_i={ x₁,x₂,…,x_MAnd Y_i={ y₁,y₂,…,y_M, M is for effectively to count out, and N is time period sum, then have

$d(X_i,Y_i)=\frac{\underset{i=1}{\overset{M}{\Σ}}{x}_i{y}_i}{\sqrt{\underset{i=1}{\overset{M}{\Σ}}{x}_i^2}\sqrt{\underset{i=1}{\overset{M}{\Σ}}{y}_i^2}}$

$d(X,Y)=\frac{\underset{i=1}{\overset{N}{\Σ}}d(X_i,Y_i)}{N}$

C2, utilize behavioral pattern distance function, the historical record of user behavior pattern is carried out DBSCAN cluster, obtains K classification, take the meansigma methods of same class behavioral pattern as class center, then this K behavioral pattern is user preferences modeling.

In described step c, Deviant Behavior is determined whether: utilize behavioral pattern distance function to calculate the similarity of user behavior pattern and user preferences modeling according to user preferences modeling, when similarity is lower than threshold value, the behavior judging user day is Deviant Behavior, then this user behavior score is directly judged to negative point.

In described step c, choose C class evaluation index, C class evaluation criterion weight in conjunction with customer service behavioral data $W=\{w_1,w_2,...,w_C\},\underset{i=1}{\overset{C}{\Σ}}{w}_i=1,,$ By $S=\underset{i=1}{\overset{C}{\Σ}}{w}_i{s}_i$ Calculate user behavior score, wherein s_iFor single item evaluation index score, draw assessment result according to user behavior score.

Described evaluation index selects user to move always distance Len, user dwell times Count, user and effectively stops number ECount, and effectively stopping number is the dwell point number having service data manipulating at stop site user.

Employing it is an advantage of the current invention that:

1, adopt after the present invention, it is not necessary to use extra work attendance to check card equipment, reduce deployment expense, be conducive to enterprise to promote the use of.

2, adopt after the present invention, it is possible to from the original mixed and disorderly track data of user, obtaining the behavioral pattern of user, accurate response goes out the Behavior law of user's every day, greatly reduce traditional work attendance and check card the possibility of cheating in violation of rules and regulations.

3, the present invention is by conjunction with user's historical behavior pattern, it is judged that user's Deviant Behavior.

4, the present invention calculates user behavior score according to selected evaluation index, sets up unified evaluation criterion, objectively responds out process of work and the work efficiency of user.

Accompanying drawing explanation

Fig. 1 is schematic flow sheet of the present invention

Fig. 2 is track data of the present invention and effective place schematic diagram

Detailed description of the invention

Embodiment 1

A kind of Trustworthy user behaviour method excavated based on place, it is characterised in that: utilize the behavioral pattern of trajectory data mining user, analyze the Behavior law of user, user behavior is estimated.

Described method specifically includes following steps:

A, acquisition positional information obtain track data, and to track data prediction;

B, by track data obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user;

C, utilize the historical record of user behavior pattern to update user preferences modeling, determine whether Deviant Behavior according to user preferences modeling, calculate user behavior score.

In described step a, mobile equipment reports track data to be L={l₁,l₂,…,l_n, wherein l_i=(lat_i,long_i,time_i) represent longitude and latitude and time, first remove and track data repeats a little, again through Kalman filter rate of filtration abnormity point, smooth track data, make track data closer to real travel path.

In described step b, adopt based on seasonal effect in time series clustering algorithm, obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user.

Described step b specifically includes:

B1, choose two parameter value place maximum magnitude D_maxWith effective place time span T;

B2, successively track data being carried out following process: when the distance of adjacent two tracing points is less than place range threshold D, two tracing points merge into a new tracing point, new tracing point participates in processing next time; When the distance of adjacent two tracing points is more than place range threshold D_maxTime, and the time span of previous tracing point more than effective place time threshold T time, this tracing point is an effective place, represent p={lat, lng, start_ts, end_ts};

B3, user count out as M effectively, within one day, are divided into N number of time period, and user behavior pattern is expressed as M × N matrix P=[f_ij], every a line represents certain effective place of user, and certain time period in one day, f are shown in each list_ijExpression rests on the probability in i-th place in the jth time period.

Described step c specifically includes:

C1, by Cosine coefficient, define user behavior pattern distance function. For behavioral pattern X and behavioral pattern Y, the place distribution vector of i-th time period is X_i={ x₁,x₂,…,x_MAnd Y_i={ y₁,y₂,…,y_M, M is for effectively to count out, and N is time period sum, then have

$d(X_i,Y_i)=\frac{\underset{i=1}{\overset{M}{\Σ}}{x}_i{y}_i}{\sqrt{\underset{i=1}{\overset{M}{\Σ}}{x}_i^2}\sqrt{\underset{i=1}{\overset{M}{\Σ}}{y}_i^2}}$

$d(X,Y)=\frac{\underset{i=1}{\overset{N}{\Σ}}d(X_i,Y_i)}{N}$

C2, utilize behavioral pattern distance function, the historical record of the behavioral pattern of user is carried out DBSCAN cluster, obtains K classification, take the meansigma methods of same class behavioral pattern as class center, then this K behavioral pattern is user preferences modeling;

C3, determine whether Deviant Behavior according to user preferences modeling: utilize behavioral pattern distance function to calculate the similarity of user behavior pattern and user preferences modeling, when similarity is lower than threshold value, the behavior judging user day is Deviant Behavior, then this user behavior score is directly judged to negative point;

C3, choose C class evaluation index, evaluation criterion weight in conjunction with customer service behavioral data $W=\{w_1,w_2,...,w_C\},\underset{i=1}{\overset{C}{\Σ}}{w}_i=1,,$ By $S=\underset{i=1}{\overset{C}{\Σ}}{w}_i{s}_i$ Calculating user behavior score, wherein si is single item evaluation index score, draws assessment result according to user behavior score.

Assessment result is, is reflected level of effort and the efficiency of user by behavior score.

Described evaluation index selects user to move always distance Len, user dwell times Count, user and effectively stops number ECount, and effectively stopping number is the dwell point number having service data manipulating at stop site user.

Embodiment 2

A kind of Trustworthy user behaviour method excavated based on place is applied in infrastructure management company management security personnel, and security personnel needs at some places specified inspection time enough every day, and needs patrol certain number of times every day.By behavior appraisal procedure, it is possible to assessed the working condition of security personnel by the mark finally obtained. The method comprises the following steps:

The mobile equipment such as a, the smart mobile phone carried with by ensuring public security obtain positional information and obtain track data, and to track data prediction;

B, the different time sections Annual distribution in different location of being ensured public security by track data acquisition, obtain the behavioral pattern of security personnel;

C, utilize the historical record of security personnel's behavioral pattern to update user preferences modeling, determine whether Deviant Behavior according to user preferences modeling, calculate user behavior score.

In described step a, mobile equipment reports track data to be L={l₁,l₂,…,l_n, wherein l_i=(lat_i,long_i,time_i) represent longitude and latitude and timestamp. First remove and track data repeats a little, again through Kalman filter rate of filtration abnormity point smooth track data, make track data closer to real travel path. Pretreated track data is as shown in the table:

Table 1: user trajectory data

In described step b, adopt based on seasonal effect in time series clustering algorithm, obtain security personnel's different time sections Annual distribution in different location, obtain the behavioral pattern of user.

Described step b specifically includes:

B1, choose place maximum magnitude be 100 meters and effectively place time span be 6 minutes;

B2, successively track data being carried out following process: when the distance of adjacent two tracing points is less than place range threshold 100 meters, two tracing points merge into a new tracing point, new tracing point participates in processing next time; When the distance of adjacent two tracing points is more than place range threshold 100 meters, and the time span of previous tracing point more than effective place time threshold 6 minutes time, this tracing point is an effective place, is expressed as p={lat, lng, start_ts, end_ts};

B3, user count out as M effectively, within one day, are divided into 24 time periods, and user behavior pattern is expressed as M × 24 matrix P=[f_ij], every a line represents certain effective place of user, and certain time period in one day, f are shown in each list_ijExpression rests on the probability in i-th place in the jth time period,One day behavioral pattern of user is as shown in the table:

Table 2: user behavior pattern

Described step c specifically includes:

C1, by Cosine coefficient, define user behavior pattern distance function. For behavioral pattern X and behavioral pattern Y, the place distribution vector of i-th time period is X_i={ x₁,x₂,…,x_MAnd Y_i={ y₁,y₂,…,y_M, M is for effectively to count out, and N is time period sum, then have

$d(X_i,Y_i)=\frac{\underset{i=1}{\overset{M}{\Σ}}{x}_i{y}_i}{\sqrt{\underset{i=1}{\overset{M}{\Σ}}{x}_i^2}\sqrt{\underset{i=1}{\overset{M}{\Σ}}{y}_i^2}}$

$d(X,Y)=\frac{\underset{i=1}{\overset{N}{\Σ}}d(X_i,Y_i)}{N}$

C2, utilize behavioral pattern distance function, the historical record of the behavioral pattern of user is carried out DBSCAN cluster, obtains K classification, take the meansigma methods of same class behavioral pattern as class center, then this K behavioral pattern is user preferences modeling;

C3, determine whether Deviant Behavior according to user preferences modeling: utilize behavioral pattern distance function to calculate the similarity of user behavior pattern and user preferences modeling, when similarity is lower than threshold value 0.3, it is judged that the behavior on user same day is Deviant Behavior;

C4, choose mobile total distance Dist, effective dwell point number ECount, cycle patrol times N period is as 3 evaluation indexes, wherein displacement Dist is the distance summation of consecutive points in track sets, and effective dwell point number ECount is the number resting on the place specified and the time of staying more than 30 minutes, and cycle patrol times N period is the number of times completing work in all appointed places, evaluation criterion weight W={0.2,0.3,0.5}, byCalculate user behavior score, wherein $s_i=\left\{\begin{array}{c}100*\frac{N_i}{{\mathrm{Normal}}_i},N_i\≤{\mathrm{Normal}}_i\\ 100,N_i>{\mathrm{Normal}}_i\end{array}\right.,$ N_iFor user's statistical value on index i, Normal_iFor the index i standard value preset.Then:

When score is more than 80, reflect that this employee completes work well.

When score is more than 60, reflects that this employee is basically completed work, but completeness is not high

When score is less than 60, reflect that this employee is not properly completed work.

Claims (9)

1. the Trustworthy user behaviour method excavated based on place, it is characterised in that: utilize the behavioral pattern of trajectory data mining user, analyze the Behavior law of user, user behavior is estimated.

2. the Trustworthy user behaviour method excavated based on place according to claim 1, it is characterised in that: described method specifically includes following steps:

A, acquisition positional information obtain track data, and to track data prediction;

B, by track data obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user;

C, utilize the historical record of user behavior pattern to update user preferences modeling, determine whether Deviant Behavior according to user preferences modeling, calculate user behavior score.

3. the Trustworthy user behaviour method excavated based on place according to claim 2, it is characterised in that: in described step a, mobile equipment reports track data to be L={l₁,l₂..., l_n, wherein l_i=(lat_i,long_i,time_i) represent longitude and latitude and time, first remove and track data repeats a little, again through Kalman filter rate of filtration abnormity point, smooth track data, make track data closer to real travel path.

4. the Trustworthy user behaviour method excavated based on place according to claim 3, it is characterized in that: in described step b, adopt based on seasonal effect in time series clustering algorithm, obtain user's different time sections Annual distribution in different location, obtain the behavioral pattern of user.

5. the Trustworthy user behaviour method excavated based on place according to claim 4, it is characterised in that: described step b specifically includes:

B1, choose two parameter value place maximum magnitude D_maxWith effective place time span T;

B2, successively track data being carried out following process: when the distance of adjacent two tracing points is less than place range threshold D, two tracing points merge into a new tracing point, new tracing point participates in processing next time; When the distance of adjacent two tracing points is more than place range threshold D_maxTime, and the time span of previous tracing point more than effective place time threshold T time, this tracing point is an effective place, is expressed as p={lat, lng, start_ts, end_ts};

B3, user count out as M effectively, within one day, are divided into N number of time period, and user behavior pattern is expressed as M × N matrix P=[f_ij], every a line represents certain effective place of user, and certain time period in one day, f are shown in each list_ijExpression rests on the probability in i-th place in the jth time period.

6. the Trustworthy user behaviour method excavated based on place according to claim 5, it is characterised in that: in described step c, user preferences modeling is drawn by following method:

C1, by Cosine coefficient, define user behavior pattern distance function. For behavioral pattern X and behavioral pattern Y, the place distribution vector of i-th time period is X_i={ x₁,x₂,...,x_MAnd Y_i={ y₁,y₂,...,y_M, M is for effectively to count out, and N is time period sum, then have

$d(X_i,Y_i)=\frac{\underset{i=1}{\overset{M}{\Σ}}{x}_i{y}_i}{\sqrt{\underset{i=1}{\overset{M}{\Σ}}{x}_i^2}\sqrt{\underset{i=1}{\overset{M}{\Σ}}{y}_i^2}}$

$d(X,Y)=\frac{\underset{i=1}{\overset{N}{\Σ}}d(X_i,Y_i)}{N}$

C2, utilize behavioral pattern distance function, the historical record of user behavior pattern is carried out DBSCAN cluster, obtains K classification, take the meansigma methods of same class behavioral pattern as class center, then this K behavioral pattern is user preferences modeling.

7. the Trustworthy user behaviour method excavated based on place according to claim 6, it is characterized in that: in described step c, Deviant Behavior is determined whether: utilize behavioral pattern distance function to calculate the similarity of user behavior pattern and user preferences modeling according to user preferences modeling, when similarity is lower than threshold value, the behavior judging user day is Deviant Behavior, then this user behavior score is directly judged to negative point.

8. the Trustworthy user behaviour method excavated based on place according to claim 7, it is characterised in that: in described step c, choose C class evaluation index, C class evaluation criterion weight W={w in conjunction with customer service behavioral data₁,w₂,…,w_C,ByCalculate user behavior score, wherein s_iFor single item evaluation index score, draw assessment result according to user behavior score.

9. the Trustworthy user behaviour method excavated based on place according to claim 8, it is characterized in that: described evaluation index selects user to move always distance Len, user dwell times Count, user and effectively stops number ECount, effectively stopping number is the dwell point number having service data manipulating at stop site user.