CN103995859A - Geographical-tag-oriented hot spot area event detection system applied to LBSN - Google Patents

Abstract

The invention discloses a geographical-tag-oriented hot spot area event detection system applied to an LBSN, and belongs to the technical field of network data processing. The detection system is operated in the LBSN and is composed of a sign-in clustering module, an area calculation module based on tag clustering and a hot-spot area event calculation module. The sign-in clustering module is used for performing clustering processing on sign-in information to obtain a geographical area corresponding to the sign-in information, a geographical tag clustering algorithm is adopted by the area calculation module based on tag clustering to obtain an in-cluster area set from the geographical area corresponding to the sign-in information, the sign-in frequency in a time window is applied by the hot spot area event calculation module to obtain a hot spot area event from the in-cluster area set, and therefore the obtained hot spot area event is provided for a user. The cluster is utilized by the geographical-tag-oriented hot spot area event detection system to perform further clustering on points in the cluster within the smaller range, and the system has the advantages of being capable of greatly reducing the calculated data volume in the LBSN and improving the calculation efficiency.

Description

A kind of hot spot region incident detection system based on geographical labels that is applied to LBSN network

Technical field

The present invention relates to a kind of technical field of registering of geographical labels, more particularly, refer to a kind of hot spot region incident detection system based on geographical labels of the LBSN of being applied to network, wherein hot spot region is the division that the cluster based on label and geographic position is carried out.

Background technology

Geographical labels (Geo Tags) refers to that the data message for describing point of interest geographic position of living in, its information content include point of interest address information, point of interest latitude and longitude information.Geographical labels is the geographical location information of digitizing point of interest better, is conducive to global metadata location and geographical location information and reviews.Geographical labels is also referred to as geographical indication.The record of registering refers to that the main body of the society is at the point of interest data message obtaining of registering.

At present, location-based social networks (LBSN Location-based Social Networking) becomes more and more popular.Due to the fast development of the 4th third generation mobile communication network fast, and to Map Services and the powerful interface support of embedded GPS module smart mobile phone, it is easily mobile subscriber and identifies their position, and shares their LBSN database.In a LBSN database, user can find and create point of interest (poi point of interest), can register in their current location, makes comments and suggestion and interpolation good friend etc.Therefore, LBSN network, as Foursquare, Facebook Places, Sina's microblogging etc., has been taked different mechanism to attract user, and has been encouraged user to share their information of registering.And, had some researchs to start to utilize that these have that user produces with the geographical labels information of registering.Because these data can allow researcher to go to analyze the interests problem of social hierarchy in the mode of data-driven, and according to the INFORMATION DISCOVERY user's that registers Move Mode, predict good friend's relation, better understand the different aspect in city.Also can utilize the information of registering to come discovering hot region.

At present the hot spot region event based on containing geographical labels finds to mainly contain a kind of mode: the first good geographic grid of artificial division, then add up the information of registering on region separately, and the total amount of registering reaches certain threshold value and is designated hot spot region.There are three problems in this method, (1) first, zoning may be divided into different grids by actual region in advance, cannot reflect actual hot spot region.(2) criterion of hot spot region be this in advance the total quantity of registering of zoning whether reached a threshold value, reach and be designated hot spot region, but do not consider the impact of time factor.(3) region that grid is divided is larger, is difficult to more accurate region, location.

Summary of the invention

For the register feature of data of LBSN, and existing hot spot region event is found the deficiency of disposal route the present invention proposes a kind of hot spot region incident detection system based on geographical labels.This hot spot region incident detection system synthesis has considered that user is at the recent historical record of registering, building coarseness region in conjunction with geospatial information divides, then, adopt the clustering algorithm of geographical labels to calculate fine granularity regional extent, finally in region, calculate the hot spot region under certain hour window.The hot spot region incident detection system based on geographical labels of the present invention's design is embedded in LBSN database, is following the LBSN network operation.

A kind of hot spot region incident detection system based on geographical labels that is applied to LBSN network of the present invention's design, arranges the described hot spot region incident detection system (3) based on geographical labels between the LBSN database (2) in described LBSN network and user (1);

The described hot spot region incident detection system (3) based on geographical labels includes the cluster module (31) of registering, region computing module (32) and hot spot region event computing module (33) based on label clustering; Described hot spot region event computing module (33) is the interface that is connected between LBSN database (2) and user (1);

The cluster of registering module (31) first aspect is for sending to LBSN database (2) the solicited message Q that registers that contains geographical labels _31-2, described Q _31-2=R_POI _p(x, y), POI;

R_POI _p(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude;

POI represents geographical labels; Any one geographical labels in described POI is designated as a, and another geographical labels is designated as b, a, b ∈ POI;

The cluster of registering module (31) second aspect is according to Q _31-2=R_POI _p(x, y), POI can search out the record of registering mating with geographical labels POI in LBSN database (2), is designated as the return message Q that registers _2-31;

The cluster of registering module (31) third aspect is to the return message Q that registers receiving _2-31carry out the processing of k-means clustering method according to cluster kcluster-span interval time, obtain region unit information Q _31-32, described Q _31-32={ regX ₁, regX ₂..., regX _y, then by Q _31-32export to the region computing module (32) based on label clustering;

RegX ₁represent first region unit in any one geographic area R;

RegX ₂represent second region unit in any one geographic area R;

RegX _yrepresent last region unit in any one geographic area R;

Y represents region unit number;

Region computing module (32) first aspect based on label clustering is for receiving area block message Q _31-32={ regX ₁, regX ₂..., regX _y;

Region computing module (32) second aspect based on label clustering according to geographical labels cluster strategy POI-CP to Q _31-32={ regX ₁, regX ₂..., regX _yprocess, obtain restraining geographical labels region unit Q _32-2; And convergence geographical labels region unit is written to LBSN database (2);

Hot spot region event computing module (33) first aspect is accepted user's (1) hot spot region inquiry request Request, described Request={Geo (x, y), dist, Hot}, and by Request={Geo (x, y), dist, Hot} is transmitted to LBSN database (2);

Hot spot region event computing module (33) second aspect is according to Request={Geo (x, y), dist, Hot} can search out (the x with Geo in LBSN database (2), y) hot spot region of coupling, is designated as inquiry return message Q _2-33;

Hot spot region event computing module (33) third aspect according to the frequency strategy POI-TP that registers under time window to described Q _2-33carry out computing, obtain region focus incident and rank ChecFreq, and described ChecFreq is fed back to user (1).

In the present invention, described geographical labels cluster strategy POI-CP has the following step:

Extraction belongs to the same area piece regX _yin the step of geographical labels POI;

Calculating belongs to the same area piece regX _yin the position number of geographical labels POI step;

Calculate the place-centric point of geographical labels POI and geographical labels position between maximum linear distance then described in judgement with zone radius threshold value r _{threshold value}size, if by r _{threshold value}assignment is to the region unit distance correlation radius in affiliated area if choose maximum linear distance as the region unit distance correlation radius in affiliated area and then by the distance correlation radius of a geographical labels distance correlation radius with b geographical labels sum is than any two geographical labels a in upper geographical labels POI, the central point distance between b obtain distance correlation ${H\_\mathrm{rel}}_{a-b}^{{\mathrm{regX}}_y}=\frac{{\mathrm{rD}}_a^{{\mathrm{regX}}_y}+{\mathrm{rD}}_b^{{\mathrm{regX}}_y}}{{\mathrm{CLD}}_{a-b}^{{\mathrm{regX}}_y}}$ Step;

Calculate any two geographical labels a in geographical labels POI, between b semantic dependency ${S\_\mathrm{rel}}_{a-b}^{Q_{31-32}}=1-\frac{E_{a-b}^{Q_{31-32}}}{\max \left(L_{a-b}^{Q_{31-32}}\right)}$ Step;

According to described with described with distance correlation threshold value rel _distance, semantic dependency threshold value rel _semanticcontrast, and merge region unit regX according to comparing result _ystep;

If and time, by the position of registering of b geographical labels merge to the position of registering of a geographical labels

If or time, the position of registering of b geographical labels not with the position of registering of a geographical labels carry out region unit merging.

In the present invention, the concrete mode of the frequency strategy POI-TP that registers under described time window is: hot spot region event computing module (33) can calculate in real time user (1) and ask any region regX _yfocus incident, register and record Q when ask separately in region to contain geography by the end of the history of current time t to focus region clustering database (2) _2-33, event description is $\mathrm{Trend}=\frac{{\mathrm{ChecFreq}}^t-{\mathrm{ChecFreq}}^{t-1}}{\mathrm{\ΔT}},$ Δ T represents time window, Δ T=|t-(t-1) |, t represents current time, t-1 represents previous moment, ChecFreq ^trepresent the frequency of registering of current time t, ChecFreq ^t-1represent the frequency of registering of previous moment t-1;

The active degree Rank of event is directly proportional with the event of continuing of registering to the frequency of registering of Δ T, that is:

$\mathrm{Rank}=\underset{i=1}{\overset{{\mathrm{regX}}_y}{\mathrm{\Σ}}}\frac{{\mathrm{ChecFreq}}_i^t}{{\mathrm{\Ω}}_i}\×\sqrt{1+\frac{\underset{{j\∈\mathrm{regX}}_j}{\max }\left({\mathrm{CU}}^j\right)}{{\mathrm{regX}}_y}};$

Wherein be illustrated in the frequency of registering in time window Δ T, regX _yrepresent any region, the element of suing for peace, i represents summing target, Ω _ibe illustrated in time window Δ T user (1) the scope inner region of the asking total quantity of registering, represent the time window number in current all hot spot regions maximal value, j represents the area identification number of time window number maximum.

The advantage of the hot spot region incident detection system based on geographical labels of the present invention's design is:

1. use hot spot region of the present invention incident detection system to solve the low defect of hot spot region precision that adopts artificial division region to cause at LBSN database.Hot spot region of the present invention incident detection system is carried out cluster to hot spot region on Label space first in the ground, obtains rough hot spot region and divides, and has reduced the data volume in cluster process.

2. hot spot region of the present invention incident detection system adopts in coarseness bunch and carries out fine-grained excavation, and the hot spot region that obtains being true to life makes the hot spot region event of obtaining find better realistic demand.

3. hot spot region of the present invention incident detection system adopts time window to divide hot spot region, goes out hot spot region by the rate of change threshold search of registering, and the query time of LBSN network is shortened, and has improved response speed.

4. hot spot region of the present invention incident detection system adopts modular design, and LBSN network by hot spot region event computing module for being connected interface, realize and user interactions, improved user and inquired about the efficiency of hot spot region event.

Brief description of the drawings

Fig. 1 is the structured flowchart that the present invention is based on the hot spot region incident detection system of geographical labels.

Fig. 2 is the sequential chart that the present invention is based on the hot spot region incident detection system of geographical labels.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail.

In Fig. 1, provide the hot spot region incident detection system architecture diagram of registering based on user according to of the present invention.The hot spot region incident detection system 3 based on geographical labels of the present invention design is set between existing LBSN database 2 and user 1, and this system includes the cluster module 31 of registering, region computing module 32 and hot spot region event computing module 33 based on label clustering.Described hot spot region event computing module 33 is the interface that is connected between LBSN database 2 and user 1.

In the present invention, the history information of registering that the LBSN database 2 that utilizes location-based social networks (LBSN Location-based Social Networking) to form provides is carried out hot spot region incident detection, is the data source of finding out hot spot region related information.

In the present invention, the point of interest POI information of registering in any one geographic area is designated as R_POI _p(x, y), POI.R represents geographic area, and POI is illustrated in the character string in R, i.e. the geographical labels of the point of interest of geographic area, place is also source file string required in semantic analysis, and the character length of POI is designated as L _pOI, p represents the number of times that the main body of the society is registered at POI, POI _p(x, y) represents the sign-in desk geographic position of the p time, and x represents longitude, and y represents latitude.

Usually, for the statement of present patent application content, geographical labels POI can be set to dining room A, the B of hospital, the C of library, the teaching building D etc. in any region R; Geographical labels POI adopts set form to be expressed as POI={A, B, and C, D}, expresses geographical labels POI for broad sense, and in described POI, any one geographical labels is designated as a, and another geographical labels is designated as b, a, b ∈ POI.Based on being described as of geographical labels:

The dining room A information of registering in any one geographic area R is designated as R_A _α(x, y), A; R represents geographic area, and A is illustrated in the character string in R, and the character length of A is designated as L _a(i.e. " geographic area " " dining room ", L _a2 bytes of=12, Chinese character), a represents the number of times that the main body of the society is registered at A, A _α(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude.

Hospital's B information of registering in any one geographic area R is designated as R_B _β(x, y), B; R represents geographic area, and B is illustrated in the character string in R, and the character length of B is designated as L _b(i.e. " geographic area " " hospital ", L _a=12), β represents the number of times that the main body of the society is registered at B, B _β(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude.

Library's C information of registering in any one geographic area R is designated as R_C _γ(x, y), C; R represents geographic area, and C is illustrated in the character string in R, and the character length of C is designated as L _c(i.e. " geographic area " " library ", L _a=14), γ represents the number of times that the main body of the society is registered at C, C _γ(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude.

The teaching building D information of registering in any one geographic area R is designated as R_D _θ(x, y), D; R represents geographic area, and D is illustrated in the character string in R, and the character length of D is designated as L _d(i.e. " geographic area " " teaching building ", L _a=14), θ represents the number of times that the main body of the society is registered at D, D _θ(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude.

User 1

User 1 is in the time using the hot spot region incident detection system 3 based on geographical labels of the present invention design, and first aspect is interest request Request={Geo (x, y), dist, and Hot} sends to hot spot region event computing module 33; Second aspect is for receiving the real-time hot spot region of the cluster information ChecFreq that hot spot region event computing module 33 returns.

Described recommendation request Request={Geo (x, y), dist, the Geo (x, y) in Hot} represents the geographic position at request user place, and x is longitude, and y is latitude; Dist represents the interest distance radius that user arranges; Hot represents the hot spot region that user pays close attention to.

Shown in Fig. 1 and Fig. 2, in the present invention, the hot spot region incident detection system 3 based on geographical labels includes the cluster module 31 of registering, region computing module 32 and hot spot region event computing module 33 based on label clustering.To describe modules below in detail:

The cluster of registering module 31

The cluster of registering module 31 first aspects are for sending the user of the containing geographical labels information Q that registers to LBSN database 2 _31-2=R_POI _p(x, y), POI;

The cluster of registering module 31 second aspects are according to Q _31-2=R_POI _p(x, y), POI can search out multiple registering of mating with geographical labels POI and record Q in LBSN database 2 _2-31;

The cluster of registering module 31 third aspect are to the information Q that registers that returns receiving _2-31carry out the processing of k-means clustering method according to cluster kcluster-span interval time, obtain region unit information Q _31-32={ regX ₁, regX2 _,, regX _y, then by Q _31-32export to the region computing module 32 based on label clustering.

In the present invention, described Q _31-2=R_POI _p(x, y), the R_POI in POI _p(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude.POI represents geographical labels, i.e. R_POI _pthe geographic name at (x, y) place is also the content that source file string is recorded; Described POI={A, B, C, D}, A is dining room geographical labels, and B is hospital's geographical labels, and C is library's geographical labels, and D is teaching building geographical labels.

In the present invention, region unit information Q _31-32={ regX ₁, regX ₂..., regX _yin regX ₁represent first region unit in any one geographic area R, regX ₂represent second region unit in any one geographic area R, regX _yrepresent last region unit in any one geographic area R, y represents region unit number.RegX _yalso referred to as any one region unit marking off in any one geographic area R.

Enumerate, meet registering of dining room A and record R_A ₁(x, y), A, R_A ₂(x, y), A, R_A ₃(x, y), A, R_A ₄(x, y), A, R_A ₅(x, y), A ..., R_A _α(x, y), A;

R_A ₁(x, y) represents first sign-in desk position of dining room A;

R_A ₂(x, y) represents second sign-in desk position of dining room A;

R_A ₃(x, y) represents the 3rd the sign-in desk position of dining room A;

R_A ₄(x, y) represents the 4th the sign-in desk position of dining room A;

R_A ₅(x, y) represents the 5th the sign-in desk position of dining room A;

R_A _α(x, y) represents last sign-in desk position of dining room A; For convenience of description, R_A _α(x, y) is also referred to as any one sign-in desk position of dining room A;

Enumerate, meet registering of the B of hospital and record R_B ₁(x, y), B, R_B ₂(x, y), B, R_B ₃(x, y), B, R_B ₄(x, y), B, R_B ₅(x, y), B ..., R_B _β(x, y), B;

R_B ₁(x, y) represents first sign-in desk position of the B of hospital;

R_B ₂(x, y) represents second sign-in desk position of the B of hospital;

R_B ₃(x, y) represents the 3rd the sign-in desk position of the B of hospital;

R_B ₄(x, y) represents the 4th the sign-in desk position of the B of hospital;

R_B ₅(x, y) represents the 5th the sign-in desk position of the B of hospital;

R_B _β(x, y) represents last sign-in desk position of the B of hospital; For convenience of description, R_B _β(x, y) is also referred to as any one sign-in desk position of the B of hospital;

Enumerate, meet registering of the C of library and record R_C ₁(x, y), C, R_C ₂(x, y), C, R_C ₃(x, y), C, R_C ₄(x, y), C, R_C ₅(x, y), C ..., R_C _γ(x, y), C;

R_C ₁(x, y) represents first sign-in desk position of the C of library;

R_C ₂(x, y) represents second sign-in desk position of the C of library;

R_C ₃(x, y) represents the 3rd the sign-in desk position of the C of library;

R_C ₄(x, y) represents the 4th the sign-in desk position of the C of library;

R_C ₅(x, y) represents the 5th the sign-in desk position of the C of library;

R_C _γ(x, y) represents last sign-in desk position of the C of library; For convenience of description, R_C _γ(x, y) is also referred to as any one sign-in desk position of the C of library;

Enumerate, meet registering of teaching building D and record R_D ₁(x, y), D, R_D ₂(x, y), D, R_D ₃(x, y), D, R_D ₄(x, y), D, R_D ₅(x, y), D ..., R_D _θ(x, y), D;

R_D ₁(x, y) represents first sign-in desk position of teaching building D;

R_D ₂(x, y) represents second sign-in desk position of teaching building D;

R_D ₃(x, y) represents the 3rd the sign-in desk position of teaching building D;

R_D ₄(x, y) represents the 4th the sign-in desk position of teaching building D;

R_D ₅(x, y) represents the 5th the sign-in desk position of teaching building D;

R_D _θ(x, y) represents last sign-in desk position of teaching building D; For convenience of description, R_D _θ(x, y) is also referred to as any one sign-in desk position of teaching building D.

According to dining room A, the B of hospital, the C of library, the teaching building D that give an example out, Q is recorded in registering of mating with geographical labels POI _2-31;

$Q_{2-31}=\left\{\begin{array}{ccccccc}R\_A:A_1(x,y)& A_2(x,y)& A_3(x,y)& A_4(x,y)& A_5(x,y)& \·\·\·& A_{\mathrm{\α}}(x,y)\\ R\_B:B_1(x,y)& B_2(x,y)& B_3(x,y)& B_4(x,y)& B_5(x,y)& \·\·\·& B_{\mathrm{\α}}(x,y)\\ R\_C:C_1(x,y)& C_2(x,y)& C_3(x,y)& C_4(x,y)& C_5(x,y)& \·\·\·& C_{\mathrm{\α}}(x,y)\\ R\_D:D_1(x,y)& D_2(x,y)& D_3(x,y)& D_4(x,y)& D_5(x,y)& \·\·\·& D_{\mathrm{\α}}(x,y)\end{array}\right\}.$

Enumerate first region unit regX ₁registering of comprising records R_A ₂(x, y), A, R_A ₃(x, y), A, R_A ₄(x, y), A, R_A ₅(x, y), A, R_A _α(x, y), A, R_B ₁(x, y), B, R_B ₂(x, y), B, R_B ₃(x, y), B, R_B ₄(x, y), B, R_C ₁(x, y), C, R_C ₂(x, y), C, R_D ₁(x, y), D and R_D _θ(x, y), D; First region unit regX ₁adopt set form to be expressed as:

${\mathrm{regX}}_1=\left\{\begin{array}{c}[R\_A_2(x,y),A],[R\_A_3(x,y),A],[R\_A_4(x,y),A],\\ [R\_A_5(x,y),A],[R\_A_{\mathrm{\α}}(x,y),A],\\ [R\_B_1(x,y),B],[R\_B_2(x,y),B],[R\_B_3(x,y),B],\\ [R\_B_4(x,y),B],\\ [R\_C_1(x,y),C],[R\_C_2(x,y),C],\\ [R\_D_1(x,y),D],[R\_D_{\mathrm{\θ}}(x,y),D]\end{array}\right\}.$

Enumerate second region unit regX ₂registering of comprising records R_A ₁(x, y), A, R_B ₅(x, y), B, R_B _β(x, y), B, R_C ₃(x, y), C, R_C ₄(x, y), C, R_C ₅(x, y), C and R_C _γ(x, y), C; Second region unit regX ₂adopt set form to be expressed as:

${\mathrm{regX}}_2=\left\{\begin{array}{c}[R\_A_1(x,y),A],\\ [R\_B_5(x,y),B],[R\_B_{\mathrm{\β}}(x,y),B],\\ [R\_C_3(x,y),C],[R\_C_4(x,y),C],[R\_C_5(x,y),C],\\ [R\_C_{\mathrm{\γ}}(x,y),C]\end{array}\right\}.$

Enumerate last region unit regX _yregistering of comprising records R_D ₂(x, y), D, R_D ₃(x, y), D, R_D ₄(x, y), D and R_D ₅(x, y), D; Last region unit regX _yadopt set form to be expressed as:

${\mathrm{regX}}_y=\left\{\begin{array}{c}[R\_D_2(x,y),D],[R\_D_3(x,y),D],[R\_D_4(x,y),D],\\ [R\_D_5(x,y),D]\end{array}\right\}.$

Record regional piece, adopting set form to express region unit information is Q _31-32={ regX ₁, regX ₂..., regX _y.

In the present invention, k-means clustering method please refer to " large data interconnection net large-scale data excavate and distributed treatment ", Wang Binyi, September in 2012 the 1st edition.

In the present invention, the cluster of registering module 31 is set to simple task, is in order to obtain the rough record of registering in the middle of the LBSN database 2 huge, by the rough record application k-means clustering method processing of registering, the record aggregate of registering can be arrived to region separately.

Use the R_POI in k-means clustering method process database _p(x, y), does not externally provide on line and processes in real time service, and the object of the cluster of registering module 31 is the scopes (being called the zoning of coarseness) that roughly mark off region CR; In addition, the scope of region CR register on space, change not obvious, can be according to the division of cluster kcluster-span interval time online lower treatment plot territory CR.

Region computing module 32 based on label clustering

Region computing module 32 first aspects based on label clustering are for receiving area block message Q _31-32={ regX ₁, regX ₂..., regX _y;

Region computing module 32 second aspects based on label clustering according to geographical labels cluster strategy POI-CP to Q _31-32={ regX ₁, regX ₂..., regX _yprocess, obtain restraining geographical labels region unit Q _32-2; And will restrain geographical labels region unit Q _32-2be written to LBSN database 2.

In the present invention, enumerate the concrete processing of explanation to geographical labels below, i.e. the implementation step of geographical labels cluster strategy POI-CP:

(1) first area piece

In the present invention, to first region unit regX ₁in the geographical labels treatment step that carries out geographical labels cluster strategy POI-CP be:

Step 101: extract the geographical labels belonging in the same area piece

From Q _31-32={ regX ₁, regX ₂..., regX _yin extract and meet first region unit regX ₁geographical labels, if first region unit regX ₁in geographical labels include A, B and C, meet regX ₁geographical labels adopt set description be;

${\mathrm{regX}}_1=\left\{\begin{array}{c}[R\_A_1(x,y),A],[R\_A_2(x,y),A],[R\_A_3(x,y),A],\\ [R\_A_4(x,y),A],[R\_A_5(x,y),A],[R\_A_{\mathrm{\α}}(x,y),A],\\ [R\_B_1(x,y),B],[R\_B_2(x,y),B],[R\_B_3(x,y),B],\\ [R\_B_4(x,y),B],\\ [R\_C_1(x,y),C],[R\_C_2(x,y),C]\end{array}\right\}.$

Step 102: the position number of geographical labels is obtained;

To first region unit regX ₁in geographical labels A carry out position classification, be met regX ₁in A geographical labels position ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_1(x,y),R\_A_2(x,y),R\_A_3(x,y),\\ R\_A_4(x,y),R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\};$ Described A is at regX ₁the number of times of middle appearance is designated as and

To first region unit regX ₁in geographical labels B carry out position classification, be met regX ₁in B geographical labels position ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\};$ Described B is at regX ₁the number of times of middle appearance is designated as and

To first region unit regX ₁in geographical labels C carry out position classification, be met regX ₁in C geographical labels position ${\mathrm{Add}}_C^{X_1}=\{R\_C_1(x,y),R\_C_2(x,y)\};$ Described C is at regX ₁the number of times of middle appearance is designated as and ${\mathrm{\α}}_C^{{\mathrm{regX}}_1}=2.$

To first region unit regX ₁in geographical labels D carry out position classification, be met regX ₁in D geographical labels position ${\mathrm{Add}}_D^{X_1}=\{R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\};$ Described D is at regX ₁the number of times of middle appearance is designated as and ${\mathrm{\α}}_D^{\mathrm{reg}{X}_1}=2.$

In the present invention, statistics is at first region unit regX ₁in the number of times of registering of all geographical labels, be designated as and ${\mathrm{\α}}_{\mathrm{POI}}^{{\mathrm{regX}}_1}={\mathrm{\α}}_A^{{\mathrm{regX}}_1}+{\mathrm{\α}}_B^{{\mathrm{regX}}_1}+{\mathrm{\α}}_C^{{\mathrm{regX}}_1}+{\mathrm{\α}}_D^{\mathrm{reg}{X}_1}.$

Step 103: distance correlation

Step 103-1: resolve ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_1(x,y),R\_A_2(x,y),R\_A_3(x,y),\\ R\_A_4(x,y),R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\}$ In longitude mean value $\stackrel{\‾}{x}\_A=\frac{\underset{i=1}{\overset{{\mathrm{\α}}^{X_1}}{\mathrm{\Σ}}}R\_A_i\left(x\right)}{{\mathrm{\α}}^{X_1}}$ With latitude mean value $\stackrel{\‾}{y}\_A=\frac{\underset{i=1}{\overset{{\mathrm{\α}}^{X_1}}{\mathrm{\Σ}}}R\_A_i\left(y\right)}{{\mathrm{\α}}^{X_1}},$ I represents the summing target in summation relation, be met regX ₁in the place-centric point of A geographical labels position resolve ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_1(x,y),R\_A_2(x,y),R\_A_3(x,y),\\ R\_A_4(x,y),R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\}$ In each label point position arrive distance, and select maximum linear distance, be designated as

Resolve ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\}$ In longitude mean value $\stackrel{\‾}{x}\_B=\frac{\underset{j=1}{\overset{{\mathrm{\β}}^{X_1}}{\mathrm{\Σ}}}R\_B_j\left(x\right)}{{\mathrm{\β}}^{X_1}}$ With latitude mean value $\stackrel{\‾}{y}\_B=\frac{\underset{j=1}{\overset{{\mathrm{\β}}^{X_1}}{\mathrm{\Σ}}}R\_B_j\left(x\right)}{{\mathrm{\β}}^{X_1}},$ J represents the summing target in summation relation, be met regX ₁in the place-centric point of B geographical labels position resolve ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\}$ In each label point position arrive distance, and select maximum linear distance and be designated as

Resolve ${\mathrm{Add}}_C^{X_1}=\{R\_C_1(x,y),R\_C_2(x,y)\}$ In longitude mean value $\stackrel{\‾}{x}\_C=\frac{\underset{m=1}{\overset{{\mathrm{\γ}}^{X_1}}{\mathrm{\Σ}}}R\_C_m\left(x\right)}{{\mathrm{\γ}}^{X_1}}$ With latitude mean value $\stackrel{\‾}{y}\_C=\frac{\underset{m=1}{\overset{{\mathrm{\γ}}^{X_1}}{\mathrm{\Σ}}}R\_C_m\left(y\right)}{{\mathrm{\γ}}^{X_1}},$ M represents the summing target in summation relation, obtain the place-centric point of AC geographical labels position resolve ${\mathrm{Add}}_C^{X_1}=\{R\_C_1(x,y),R\_C_2(x,y)\}$ In each label point position arrive distance, and select maximum linear distance and be designated as

Resolve ${\mathrm{Add}}_D^{X_1}=\{R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\}$ In longitude mean value $\stackrel{\‾}{x}\_D=\frac{\underset{n=1}{\overset{{\mathrm{\θ}}^{X_1}}{\mathrm{\Σ}}}R\_C_n\left(x\right)}{{\mathrm{\θ}}^{X_1}}$ With latitude mean value $\stackrel{\‾}{y}\_D=\frac{\underset{n=1}{\overset{{\mathrm{\θ}}^{X_1}}{\mathrm{\Σ}}}R\_C_n\left(y\right)}{{\mathrm{\θ}}^{X_1}},$ N represents the summing target in summation relation, obtain the place-centric point of AD geographical labels position resolve ${\mathrm{Add}}_D^{X_1}=\{R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\}$ In each label point position arrive distance, and select maximum linear distance and be designated as

In the present invention, statistics is at first region unit regX ₁in all geographical labels and the place-centric point of geographical labels position between maximum linear distance and ${\mathrm{LD}}_{\mathrm{POI}-\max }^{\mathrm{reg}{X}_1}=\{{\mathrm{LD}}_{A-\max }^{X_1},{\mathrm{LD}}_{B-\max }^{X_1},{\mathrm{LD}}_{C-\max }^{X_1},{\mathrm{LD}}_{D-\max }^{X_1}\}.$

Step 103-2: setting area piece radius threshold value is designated as r _{threshold value};

If maximum linear distance is less than zone radius threshold value r _{threshold value}, by r _{threshold value}assignment is to the region unit distance correlation radius in affiliated area

If maximum linear distance is more than or equal to zone radius threshold value r _{threshold value}, choose maximum linear distance as the region unit distance correlation radius in affiliated area

In like manner can obtain: the distance correlation radius of a geographical labels is designated as the distance correlation radius of b geographical labels is designated as

Enumerate, if by r _{threshold value}assignment is given and is met regX ₁in region unit distance correlation radius if will assignment is given and is met regX ₁in region unit distance correlation radius

Enumerate, if by r _{threshold value}assignment is given and is met regX ₁in region unit distance correlation radius if will assignment is given and is met regX ₁in region unit distance correlation radius

Enumerate, if by r _{threshold value}assignment is given and is met regX ₁in region unit distance correlation radius if will assignment is given and is met regX ₁in region unit distance correlation radius

Enumerate, if by r _{threshold value}assignment is given and is met regX ₁in region unit distance correlation radius if will assignment is given and is met regX ₁in region unit distance correlation radius

Step 103-3: calculate and meet regX ₁in the central point distance of any two geographical labels positions;

with central point distance be designated as

with central point distance be designated as

with central point distance be designated as

with central point distance be designated as

with central point distance be designated as

In the present invention, statistics is at first region unit regX ₁in geographical labels POI in any two geographical labels a, central point between b distance is designated as

Step 103-4: definition meets regX ₁in distance correlation

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{A-B}^{X_1}=\frac{r{D}_A^{X_1}+r{D}_B^{X_1}}{{\mathrm{CLD}}_{A-B}^{\mathrm{reg}{X}_1}};$

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{A-C}^{X_1}=\frac{r{D}_A^{X_1}+r{D}_C^{X_1}}{{\mathrm{CLD}}_{A-C}^{\mathrm{reg}{X}_1}};$

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{A-D}^{X_1}=\frac{r{D}_A^{X_1}+r{D}_D^{X_1}}{{\mathrm{CLD}}_{A-D}^{\mathrm{reg}{X}_1}};$

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{B-C}^{X_1}=\frac{r{D}_B^{X_1}+r{D}_C^{X_1}}{{\mathrm{CLD}}_{B-C}^{\mathrm{reg}{X}_1}}.$

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{B-D}^{X_1}=\frac{r{D}_B^{X_1}+r{D}_D^{X_1}}{{\mathrm{CLD}}_{B-D}^{\mathrm{reg}{X}_1}}.$

In the present invention, plan range is that known range formula calculates, as $\left|\mathrm{AB}\right|=\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2}.$

In the present invention, statistics is at first region unit regX ₁in distance correlation be designated as $H\_{\mathrm{rel}}_{a-b}^{\mathrm{reg}{X}_1}=\frac{r{D}_a^{\mathrm{reg}{X}_1}+r{D}_b^{\mathrm{reg}{X}_1}}{{\mathrm{CLD}}_{a-b}^{\mathrm{reg}{X}_1}}.$

Step 104: semantic dependency

In the present invention, definition semantic distance is the editing distance between POI, i.e. E _pOI.Described editing distance E _pOIfor source file string POI={A, B, C, D} converts target strings TPOI={A to, B, C, the sequence of operation that D} cost is minimum.The calculating of editing distance please refer to " introduction to algorithms " 218-219 page of the 1st edition the 14th printing Dec in 2009, (U.S.) Thomas H.Cormen Charles E.Leiserson Ronald L.Rivest Clifford Stein work, Pan Jingui, Gu Tiecheng, Li Chengfa, Ye Maoyi.

Enumerate, at first region unit regX ₁in the string length of dining room A be designated as the string length of the B of hospital is designated as the string length of the C of library is designated as the string length of teaching building D is designated as

Enumerate, at first region unit regX ₁in dining room A and the editing distance of the B of hospital be designated as the editing distance of dining room A and the C of library is designated as the editing distance of dining room A and teaching building D is designated as the editing distance of the B of hospital and the C of library is the editing distance of the B of hospital and teaching building D is designated as

Enumerate, at first region unit regX ₁in dining room A and the semantic dependency of the B of hospital be designated as $S\_{\mathrm{rel}}_{A-B}^{X_1}=1-\frac{E_{A-B}^{X_1}}{\max (L_A^{X_1},L_B^{X_1})};$

Enumerate, at first region unit regX ₁in dining room A and the semantic dependency of the C of library be designated as $S\_{\mathrm{rel}}_{A-C}^{X_1}=1-\frac{E_{A-C}^{X_1}}{\max (L_A^{X_1},L_C^{X_1})};$

Enumerate, at first region unit regX ₁in dining room A and the semantic dependency of teaching building D be designated as $S\_{\mathrm{rel}}_{A-D}^{X_1}=1-\frac{E_{A-D}^{X_1}}{\max (L_A^{X_1},L_D^{X_1})};$

Enumerate, at first region unit regX ₁in the B of hospital and the semantic dependency of the C of library be designated as $S\_{\mathrm{rel}}_{B-C}^{X_1}=1-\frac{E_{B-C}^{X_1}}{\max (L_B^{X_1},L_C^{X_1})}.$

Enumerate, at first region unit regX ₁in the B of hospital and the semantic dependency of teaching building D be designated as $S\_{\mathrm{rel}}_{B-D}^{X_1}=1-\frac{E_{B-D}^{X_1}}{\max (L_B^{X_1},L_D^{X_1})}.$

In the present invention, at area information Q _31-32in the semantic dependency of any two geographical labels be designated as $S\_{\mathrm{rel}}_{a-b}^{Q_{31-32}}=1-\frac{E_{a-b}^{Q_{31-32}}}{\max \left(L_{a-b}^{Q_{31-32}}\right)}.$

Step 105: whether region unit merges

Distance correlation threshold value is set and is designated as rel _distance, semantic dependency threshold value is designated as rel _semantic, and according to rel _distanceand rel _semanticwhether carry out the merging processing of region unit;

Step 105-1: if and time, will ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\}$ Merge to ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_1(x,y),R\_A_2(x,y),R\_A_3(x,y),\\ R\_A_4(x,y),R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\}$ In, be updated to ${\mathrm{Add}}_{A\_\mathrm{new}}^{X_1}=\left\{\begin{array}{c}R\_A_2(x,y),R\_A_3(x,y),R\_A_4(x,y),\\ R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y),\\ R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\};$ And will export to LBSN database, execution step 105-2;

Step 105-2: if or time, with do not carry out region unit merging; And will with export to LBSN database, execution step 105-3;

Step 105-3: if and time, will ${\mathrm{Add}}_C^{X_1}=\{R\_C_1(x,y),R\_C_2(x,y)\}$ Merge to ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_2(x,y),R\_A_3(x,y),R\_A_4(x,y),\\ R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\}$ In, be updated to ${\mathrm{Add}}_{A\_\mathrm{new}}^{X_1}=\left\{\begin{array}{c}R\_A_2(x,y),R\_A_3(x,y),R\_A_4(x,y),\\ R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y),\\ R\_C_1(x,y),R\_C_2(x,y)\end{array}\right\};$ And will export to LBSN database, execution step 105-4;

Step 105-4: if or time, with do not carry out region unit merging; And will with export to LBSN database, execution step 105-5;

If step 105-5 and time, will ${\mathrm{Add}}_D^{X_1}=\{R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\}$ Merge to ${\mathrm{Add}}_A^{X_1}=\left\{\begin{array}{c}R\_A_2(x,y),R\_A_3(x,y),R\_A_4(x,y),\\ R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y)\end{array}\right\}$ In, be updated to ${\mathrm{Add}}_{A\_\mathrm{new}}^{X_1}=\left\{\begin{array}{c}R\_A_2(x,y),R\_A_3(x,y),R\_A_4(x,y),\\ R\_A_5(x,y),R\_A_{\mathrm{\α}}(x,y),\\ R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\end{array}\right\};$ And will

Step 105-6: if or time, with do not carry out region unit merging; And will with export to LBSN database, execution step 105-7;

Step 105-7: if and time, will ${\mathrm{Add}}_C^{X_1}=\{R\_C_1(x,y),R\_C_2(x,y)\}$ Merge to ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\}$ In, be updated to ${\mathrm{Add}}_{B\_\mathrm{new}}^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y),\\ R\_C_1(x,y),R\_C_2(x,y)\end{array}\right\};$ And will export to LBSN database, execution step 105-8;

Step 105-8: if or time, with do not carry out region unit merging; And will with export to LBSN database, execution step 105-9;

Step 105-9: if and time, will ${\mathrm{Add}}_D^{X_1}=\{R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\}$ Merge to ${\mathrm{Add}}_B^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y)\end{array}\right\}$ In, be updated to ${\mathrm{Add}}_{B\_\mathrm{new}}^{X_1}=\left\{\begin{array}{c}R\_B_1(x,y),R\_B_2(x,y),R\_B_3(x,y),\\ R\_B_4(x,y),\\ R\_D_1(x,y),R\_D_{\mathrm{\θ}}(x,y)\end{array}\right\};$ And will export to LBSN database, execution step 105-10;

Step 105-10: if or time, with do not carry out region unit merging; And will with export to LBSN database.

In the present invention, by first region unit regX ₁the sign-in desk positional information that writes LBSN database 2 adopt set form to be expressed as $Q_{2-33}=\left\{{\mathrm{checkin}}_{{\mathrm{regX}}_1}\right\}.$

(2) second area piece

In the present invention, to second region unit regX ₂in the geographical labels treatment step that carries out geographical labels cluster strategy POI-CP be:

Step 201: extract the geographical labels belonging in the same area piece

From Q _31-32={ regX ₁, regX ₂..., regX _yin extract meet second region unit regX ₂geographical labels, if second region unit regX ₂in geographical labels include B and C, meet regX ₂geographical labels adopt set description be;

${\mathrm{regX}}_2=\left\{\begin{array}{c}[R\_B_5(x,y),B],[R\_B_{\mathrm{\β}}(x,y),B],\\ [R\_C_3(x,y),C],[R\_C_4(x,y),C],[R\_C_5(x,y),C],\\ [R\_C_{\mathrm{\γ}}(x,y),C]\end{array}\right\}.$

Step 202: the position number of geographical labels is obtained;

To second region unit regX ₂in geographical labels B carry out position classification, be met regX ₂in B geographical labels position ${\mathrm{Add}}_B^{X_2}=\{R\_B_5(x,y),R\_B_{\mathrm{\β}}(x,y)\};$ Described B is at regX ₂the number of times of middle appearance is designated as and ${\mathrm{\α}}_B^{{\mathrm{regX}}_2}=2.$

To second region unit regX ₂in geographical labels C carry out position classification, be met regX ₂in C geographical labels position ${\mathrm{Add}}_C^{X_2}=\left\{\begin{array}{c}R\_C_3(x,y),R\_C_4(x,y),R\_C_5(x,y),\\ R\_C_{\mathrm{\γ}}(x,y)\end{array}\right\};$ Described C is at regX ₂the number of times of middle appearance is designated as and

In the present invention, statistics is at second region unit regX ₂in the number of times of registering of all geographical labels, be designated as and ${\mathrm{\α}}_{\mathrm{POI}}^{{\mathrm{regX}}_2}={\mathrm{\α}}_B^{\mathrm{reg}{X}_2}+{\mathrm{\α}}_C^{{\mathrm{regX}}_2}.$

Step 203: distance correlation

Step 203-1: resolve ${\mathrm{Add}}_B^{X_2}=\{R\_B_5(x,y),R\_B_{\mathrm{\β}}(x,y)\}$ In longitude mean value $\stackrel{\‾}{x}\_B=\frac{\underset{j=1}{\overset{{\mathrm{\β}}^{X_2}}{\mathrm{\Σ}}}R\_B_j\left(x\right)}{{\mathrm{\β}}^{X_2}}$ With latitude mean value $\stackrel{\‾}{y}\_B=\frac{\underset{j=1}{\overset{{\mathrm{\β}}^{X_2}}{\mathrm{\Σ}}}R\_B_j\left(x\right)}{{\mathrm{\β}}^{X_2}},$ J represents the summing target in summation relation, be met regX ₂in the place-centric point of B geographical labels position resolve ${\mathrm{Add}}_B^{X_2}=\{R\_B_5(x,y),R\_B_{\mathrm{\β}}(x,y)\}$ In each label point position arrive distance, and select maximum linear distance, be designated as

Resolve ${\mathrm{Add}}_C^{X_2}=\left\{\begin{array}{c}R\_C_3(x,y),R\_C_4(x,y),R\_C_5(x,y),\\ R\_C_{\mathrm{\γ}}(x,y)\end{array}\right\}$ In longitude mean value $\stackrel{\‾}{x}\_C=\frac{\underset{m=1}{\overset{{\mathrm{\γ}}^{X_2}}{\mathrm{\Σ}}}R\_C_m\left(x\right)}{{\mathrm{\γ}}^{X_2}}$ With latitude mean value $\stackrel{\‾}{y}\_C=\frac{\underset{m=1}{\overset{{\mathrm{\γ}}^{X_2}}{\mathrm{\Σ}}}R\_C_m\left(y\right)}{{\mathrm{\γ}}^{X_2}},$ M represents the summing target in summation relation, and m ∈ γ ', is met regX ₂in the place-centric point of C geographical labels position resolve ${\mathrm{Add}}_C^{X_2}=\left\{\begin{array}{c}R\_C_3(x,y),R\_C_4(x,y),R\_C_5(x,y),\\ R\_C_{\mathrm{\γ}}(x,y)\end{array}\right\}$ In each label point position arrive distance, and select maximum linear distance, be designated as

In the present invention, statistics is at second region unit regX ₂in all geographical labels and the place-centric point of geographical labels position between maximum linear distance and ${\mathrm{LD}}_{\mathrm{POI}-\max }^{\mathrm{reg}{X}_2}=\{{\mathrm{LD}}_{B-\max }^{X_1},{\mathrm{LD}}_{C-\max }^{X_1}\}.$

Step 203-2: setting area piece radius threshold value is designated as r _{threshold value};

If maximum linear distance is less than zone radius threshold value r _{threshold value}, by r _{threshold value}assignment is to the region unit distance correlation radius r D in affiliated area _pOI;

If maximum linear distance is more than or equal to zone radius threshold value r _{threshold value}, choose maximum linear distance for the region unit distance correlation radius r D in affiliated area _pOI;

In like manner can obtain: the distance correlation radius of a geographical labels is designated as the distance correlation radius of b geographical labels is designated as

Enumerate, if by r _{threshold value}assignment is given and is met regX ₂in region unit distance correlation radius if will assignment is given and is met regX ₂in region unit distance correlation radius

Enumerate, if by r _{threshold value}assignment is given and is met regX ₂in region unit distance correlation radius if will assignment is given and is met regX ₂in region unit distance correlation radius

Step 203-3: calculate and meet regX ₂in the central point distance of any two geographical labels positions;

with central point distance be designated as

Step 203-4: definition meets regX ₂in distance correlation

Enumerate, with distance correlation be designated as $H\_{\mathrm{rel}}_{B-C}^{X_2}=\frac{r{D}_B^{X_2}+r{D}_C^{X_2}}{\mathrm{CL}{D}_{B-C}^{\mathrm{reg}{X}_2}}.$

In the present invention, statistics is at second region unit regX ₂in distance correlation be designated as $H\_{\mathrm{rel}}_{a-b}^{{\mathrm{regX}}_2}=\frac{r{D}_a^{\mathrm{reg}{X}_2}+r{D}_b^{{\mathrm{regX}}_2}}{\mathrm{CL}{D}_{a-b}^{\mathrm{reg}{X}_2}}.$

Step 204: semantic dependency

Enumerate, at second region unit regX ₂in the string length of the B of hospital be designated as the string length of the C of library is designated as

Enumerate, at second region unit regX ₂the editing distance of the B of institute of traditional Chinese medicine and the C of library is

Enumerate, at second region unit regX ₂in the B of hospital and the semantic dependency of the C of library be designated as $S\_\mathrm{re}{l}_{B-C}^{X_2}=1-\frac{E_{B-C}^{X_2}}{\max (L_B^{X_2},L_C^{X_2})}.$

In the present invention, at area information Q _31-32in the semantic dependency of geographical labels be designated as $S\_\mathrm{re}{l}_{\mathrm{POI}}^{Q_{31-32}}=1-\frac{E_{\mathrm{POI}}^{Q_{31-32}}}{\max \left(L_{\mathrm{POI}}^{Q_{31-32}}\right)}.$

Step 205: whether region unit merges

Distance correlation threshold value is set and is designated as rel _distance, semantic dependency threshold value is designated as rel _semantic, and according to rel _distanceand rel _semanticwhether carry out the merging processing of region unit;

Step 205-1: if and time, will ${\mathrm{Add}}_C^{X_2}=\left\{\begin{array}{c}R\_C_3(x,y),R\_C_4(x,y),R\_C_5(x,y),\\ R\_C_{\mathrm{\γ}}(x,y)\end{array}\right\}$ Merge to ${\mathrm{Add}}_B^{X_2}=\{R\_B_5(x,y),R\_B_{\mathrm{\β}}(x,y)\}$ In, be updated to ${\mathrm{Add}}_{B\_\mathrm{new}}^{X_2}=\left\{\begin{array}{c}R\_B_5(x,y),R\_B_{\mathrm{\β}}(x,y),\\ R\_C_3(x,y),R\_C_4(x,y),R\_C_5(x,y),\\ R\_C_{\mathrm{\γ}}(x,y)\end{array}\right\};$ And will export to LBSN database, execution step 205-2;

Step 205-2: if or time, with do not carry out region unit merging, and will with export to LBSN database.

In the present invention, by second region unit regX ₂the sign-in desk positional information that writes LBSN database 2 adopt set form to be expressed as $Q_{2-33}=\left\{{\mathrm{checkin}}_{{\mathrm{regX}}_2}\right\}.$

(3) the 3rd region units

In the present invention, to last region unit regX _ycarrying out the treatment step that geographical labels carries out geographical labels cluster strategy POI-CP is:

Step 301: extract the geographical labels belonging in the same area piece

From Q _31-32={ regX ₁, regX ₂..., regX _yin extract and meet last region unit regX _ygeographical labels, if last region unit regX _yin geographical labels be only D, meet regX _ygeographical labels be;

${\mathrm{regX}}_y\left\{\begin{array}{c}[R\_D_2(x,y),D],[R\_D_3(x,y),D],[R\_D_4(x,y),D],\\ [R\_D_5(x,y),D]\end{array}\right\}.$

Step 302: the position number of geographical labels is obtained;

To last region unit regX _yin geographical labels D carry out position classification, be met regX _yin D geographical labels position ${\mathrm{Add}}_D^{X_y}=\left\{\begin{array}{c}R\_D_2(x,y),R\_D_3(x,y),R\_D_4(x,y),\\ R\_D_5(x,y)\end{array}\right\};$ Described D is at regX _ythe number of times of middle appearance is designated as and

Step 303: distance correlation

Step 303-1: resolve ${\mathrm{Add}}_D^{x_y}=\left\{\begin{array}{c}R\_D_2(x,y),R\_D_3(x,y),R\_D_4(x,y),\\ R\_D_5(x,y)\end{array}\right\}$ In longitude mean value $\stackrel{\‾}{x}\_A=\frac{\underset{i=1}{\overset{{\mathrm{\α}}^{X_y}}{\mathrm{\Σ}}}R\_A_i\left(x\right)}{{\mathrm{\α}}^{X_y}}$ With latitude mean value $\stackrel{\‾}{y}\_A=\frac{\underset{i=1}{\overset{{\mathrm{\α}}^{X_y}}{\mathrm{\Σ}}}R\_A_i\left(y\right)}{{\mathrm{\α}}^{X_y}},$ I represents the summing target in summation relation, be met regX _yin the place-centric point of D geographical labels position resolve ${\mathrm{Add}}_D^{x_y}=\left\{\begin{array}{c}R\_D_2(x,y),R\_D_3(x,y),R\_D_4(x,y),\\ R\_D_5(x,y)\end{array}\right\}$ In each label point position arrive distance, and select maximum linear distance, be designated as

In the present invention, an in the end region unit regX of statistics _yin all geographical labels and the place-centric point of geographical labels position between maximum linear distance and ${\mathrm{LD}}_{\mathrm{POI}-\max }^{{\mathrm{regX}}_y}=\left\{{\mathrm{LD}}_{D-\max }^{X_y}\right\}.$

Step 303-2: setting area piece radius threshold value is designated as r _{threshold value};

If maximum linear distance is less than zone radius threshold value r _{threshold value}, by r _{threshold value}assignment is to the region unit distance correlation radius r D in affiliated area _pOI;

If maximum linear distance is more than or equal to zone radius threshold value r _{threshold value}, choose maximum linear distance as the region unit distance correlation radius r D in affiliated area _pOI;

Enumerate, if by r _{threshold value}assignment is given and is met regX _yin region unit distance correlation radius if will assignment is given and is met regX _yin region unit distance correlation radius

Step 303-3: calculate and meet regX _yin the central point distance of any two geographical labels positions;

Owing to meeting regX _yin geographical labels only have D, therefore r _{threshold value}centered by some distance be designated as

Step 303-4: definition meets regX _yin distance correlation

Enumerate, will distance correlation be designated as

Step 304: semantic dependency

Enumerate, in the end a region unit regX _yin the string length of teaching building D be designated as

Enumerate, in the end a region unit regX _yin owing to only having teaching building D, therefore the editing distance of D is designated as and $E_{D-0}^{X_y}=0.$

Enumerate, in the end a region unit regX _yin the semantic dependency of teaching building D be designated as $S\_\mathrm{re}{l}_{D-0}^{X_y}=1-\frac{E_{D-0}^{X_y}}{\max (L_D^{X_y},0)},$ And $S\_{\mathrm{rel}}_{D-0}^{X_y}=1.$

Step 305: whether region unit merges

Distance correlation threshold value is set and is designated as rel _distance, semantic dependency threshold value is designated as rel _semantic, and according to rel _distanceand rel _semanticwhether carry out the merging processing of region unit;

In the present invention, due to $E_{D-0}^{X_y}=0$ With $S\_{\mathrm{rel}}_{D-0}^{X_y}=1,$ Therefore with therefore last region unit regX _ydo not need to carry out region merging.

In the present invention, by last region unit regX _ythe sign-in desk positional information that writes LBSN database 2 adopt set form to be expressed as $Q_{2-33}=\left\{{\mathrm{checkin}}_{{\mathrm{regX}}_y}\right\}.$

Hot spot region event computing module 33

Hot spot region event computing module 33 first aspects are accepted user 1 hot spot region inquiry request Request, described Request={Geo (x, y), dist, Hot}, and by Request={Geo (x, y), dist, Hot} is transmitted to LBSN database 2;

Hot spot region event computing module 33 second aspects are according to Request={Geo (x, y), dist, and Hot} can search out the hot spot region with Geo (x, y) coupling in LBSN database 2, is designated as inquiry return message Q _2-33;

Hot spot region event computing module 33 third aspect according to the frequency strategy POI-TP that registers under time window to described Q _2-33carry out computing, obtain region focus incident and rank ChecFreq, and described ChecFreq is fed back to user 1.

In the present invention, hot spot region event computing module 33 can calculate user 1 in real time asks the focus incident of scope inner region, and specifically asking separately in region by the end of the request moment to focus region clustering database 2 (is that current time history t) contains the geography record of registering $Q_{2-33}=\{{\mathrm{checkin}}_{{\mathrm{regX}}_1},{\mathrm{checkin}}_{{\mathrm{regX}}_2},\·\·\·,{\mathrm{checkin}}_{\mathrm{reg}{X}_y}\}.$

In the present invention, with any region regX _yregistering according to historical geography, to detect focus incident be example to label, and definition time window is Δ T, and the frequency of registering occurring within the described Δ T time period is ChecFreq.The number of times that the described frequency ChecFreq that registers refers to that register in identical hot spot region in Δ T.Described Δ T=|t-(t-1) |, current time is t, previous moment is t-1.

In the present invention, event definition is any region regX _yin the register variable quantity of frequency ChecFreq, event description expression-form is Trend:

$\mathrm{Trend}=\frac{{\mathrm{ChecFreq}}^t-{\mathrm{ChecFreq}}^{t-1}}{\mathrm{\ΔT}}$

When Trend exceeds certain event detection threshold value Trend _{threshold value}time, hot spot region event computing module 33 is by region regX _ybe labeled as focus incident.And be surge type event flag or suddenly fall type according to Trend value positive and negative.

In the present invention, focus incident is carried out to ranking, the ranking that focus event time window number is more is front.At any one regX _yin, have under continuous time window Δ T, if the frequency ChecFreq that registers of focus incident exceedes the frequency threshold value ChecFreq that registers _{threshold value}, choose the focus incident lasting time window number continuously that exceedes threshold value, be designated as in like manner known, in all hot spot regions, focus incident is lasting time window number continuously, is designated as $\mathrm{RCU}=\{{\mathrm{CU}}^{{\mathrm{regX}}_1},{\mathrm{CU}}^{{\mathrm{regX}}_2},\·\·\·,{\mathrm{CU}}^{{\mathrm{regX}}_y}\}.$

The active degree Rank of event is directly proportional with the event of continuing of registering to the frequency of registering of Δ T, that is:

$\mathrm{Rank}=\underset{i=1}{\overset{{\mathrm{regX}}_y}{\mathrm{\Σ}}}\frac{{\mathrm{ChecFreq}}_i^t}{{\mathrm{\Ω}}_i}\×\sqrt{1+\frac{\underset{j\∈{\mathrm{regX}}_y}{\max }\left({\mathrm{CU}}^j\right)}{{\mathrm{regX}}_y}}$

Wherein be illustrated in the quantity of registering in time window Δ T, regX _yrepresent any region, the element of suing for peace, i represents summing target, Ω _ibe illustrated in time window Δ T the user 1 scope inner region of the asking total quantity of registering. represent the time window number in current all hot spot regions maximal value, j represents the area identification number of time window number maximum.

A kind of hot spot region incident detection system of registering based on geographical labels that the present invention proposes, this system belongs to the event detection technical field in location-based social network.First system always the uninterrupted LBSN that must move crawl module, in warp-wise LBSN database, write the record of registering that contains geographical labels, next can adopt the clustering algorithm of registering to obtain general regional cluster; Then the region clustering module of system can adopt the region clustering algorithm that contains geographical labels to calculate precise region; The request of hot spot region computing module meeting relative users, this is the unique service interface externally providing of system, first the query argument that this module can be submitted to according to user, submit corresponding inquiry to database, and carry out region focus incident detection algorithm to returning to the data of coming, calculate the rank of focus incident and event, and return to request user.

embodiment

Shown in Fig. 1, Fig. 2, the user who supposes to meet geographical labels in the LBSN database information Q that registers _31-2=R_POI _p(x, y), the Search Results of POI has multiple, applies cluster kcluster-span interval time and carries out the processing of k-means clustering method, obtains region unit information Q _31-32={ regX ₁, regX ₂..., regX _y.Described Q _31-2=R_POI _p(x, y), POI is raw readings, does not also carry out cluster.

Suppose Q _31-32={ regX ₁, regX ₂..., regX _ycoarseness geographic area after cluster is 6, region unit is counted y=6, i.e. Q _31-32={ regX ₁, regX ₂, regX ₃, regX ₄, regX ₅, regX ₆.

At Q _31-32in, interregional distance relation is as following table:

While supposing that request user sends local hot spot region inquiry request, the operational process of system is as follows:

Step 1: request user's recommendation request information is Request={Geo (x, y), dist, Hot}={inR ₃, 2000, Hot}, inR ₃for the location point of request, it and other regional distance are:

Can see, within the scope of request, having 3 regions to satisfy condition, i.e. regX ₁, regX ₃, regX ₄.First user's request can be received by the hot spot region event computing module 33 of hot spot region incident detection system.

Step 2: the parametric configuration that hot spot region event computing module 33 passes into user becomes rational query statement requested database, database is by these three region regX that meet the demands ₁, regX ₃, regX ₄with their history with the geographical labels record of registering $Q_{2-33}=\{{\mathrm{checkin}}_{{\mathrm{regX}}_1},{\mathrm{checkin}}_{{\mathrm{regX}}_2},\·\·\·,{\mathrm{checkin}}_{\mathrm{reg}{X}_y}\}$ Return to hot spot region event computing module 33.

Step 3: hot spot region event computing module 33 obtains registering after record, starts to calculate.The value of Δ T is 1 hour, the frequency of registering threshold value ChecFreq _{threshold value}be 100, event detection threshold value Trend _{threshold value}be 50/h, find forward 4 time windows, first time window T ₁=t-3 Δ T, second time window T ₂=t-2 Δ T, the 3rd time window T ₃=t-Δ T, the 4th time window T ₄=t.

At region regX ₁in, the frequency ChecFreq that registers of a certain event is:

Period	T 1	T 2	T 3	T 4
					ChecFreq	167	101	150	50

Its maximum is enlivened time window value continuously, and to be 3 (be focus incident lasting time window number continuously ${\mathrm{CU}}^{{\mathrm{regX}}_1}=3$ )。

At region regX ₃in, the frequency ChecFreq that registers of a certain event is:

Period	T 1	T 2	T 3	T 4
					ChecFreq	24	30	50	99

It is 3 that its maximum is enlivened time window value continuously.

At region regX ₄in, the frequency ChecFreq that registers of current period is:

Period	T 1	T 2	T 3	T 4
					ChecFreq	112	22	23	12

It is 1 that its maximum is enlivened time window value continuously.

According to ${\mathrm{Trend}}_{{\mathrm{regX}}_1}=\frac{{\mathrm{ChecFreq}}^t-{\mathrm{ChecFreq}}^{t-1}}{\mathrm{\ΔT}}=66,$ Exceed event detection threshold; do not reach event detection threshold value; exceed event detection threshold.Can see and have two hot spot regions.Respectively regX ₁and regX ₄.

For regX ₁, its time window value of enlivening is continuously 3, the active degree rank of its event:

${\mathrm{Rank}}_{{\mathrm{regX}}_1}=\underset{i=1}{\overset{4}{\mathrm{\Σ}}}\frac{{\mathrm{ChecFreq}}_i}{{\mathrm{\Ω}}_i}\×\sqrt{1+\frac{3}{4}}=0.89\×1.48=1.31.$

For regX ₄, its time window value of enlivening is continuously 1, the active degree rank of its event:

${\mathrm{Rank}}_{{\mathrm{regX}}_4}=\underset{i=1}{\overset{4}{\mathrm{\Σ}}}\frac{{\mathrm{ChecFreq}}_i}{{\mathrm{\Ω}}_i}\×\sqrt{1+\frac{1}{4}}=0.66\×1.11=0.73.$

Result of calculation is returned to user by hot spot region event computing module 33.Be that ChecFreq returns results as ChecFreq={regX ₁: 1, regX ₄: 2}.

Claims (3)

1. the hot spot region incident detection system based on geographical labels that is applied to LBSN network, is characterized in that: the described hot spot region incident detection system (3) based on geographical labels is set between the LBSN database (2) in described LBSN network and user (1);

The described hot spot region incident detection system (3) based on geographical labels includes the cluster module (31) of registering, region computing module (32) and hot spot region event computing module (33) based on label clustering; Described hot spot region event computing module (33) is the interface that is connected between LBSN database (2) and user (1);

The cluster of registering module (31) first aspect is for sending to LBSN database (2) the solicited message Q that registers that contains geographical labels _31-2, described Q _31-2=R_POI _p(x, y), POI;

R_POI _p(x, y) represents sign-in desk geographic position, and x represents longitude, and y represents latitude;

POI represents geographical labels; Any one geographical labels in described POI is designated as a, and another geographical labels is designated as b, a, b ∈ POI;

The cluster of registering module (31) second aspect is according to Q _31-2=R_POI _p(x, y), POI can search out the record of registering mating with geographical labels POI in LBSN database (2), is designated as the return message Q that registers _2-31;

The cluster of registering module (31) third aspect is to the return message Q that registers receiving _2-31carry out the processing of k-means clustering method according to cluster kcluster-span interval time, obtain region unit information Q _31-32, described Q _31-32={ regX ₁, regX ₂..., regX _y, then by Q _31-32export to the region computing module (32) based on label clustering;

RegX ₁represent first region unit in any one geographic area R;

RegX ₂represent second region unit in any one geographic area R;

RegX _yrepresent last region unit in any one geographic area R;

Y represents region unit number;

Region computing module (32) first aspect based on label clustering is for receiving area block message Q _31-32={ regX ₁, regX ₂..., regX _y;

Region computing module (32) second aspect based on label clustering according to geographical labels cluster strategy POI-CP to Q _31-32={ regX ₁, regX ₂..., regX _yprocess, obtain restraining geographical labels region unit Q _32-2; And convergence geographical labels region unit is written to LBSN database (2);

Hot spot region event computing module (33) first aspect is accepted user's (1) hot spot region inquiry request Request, described Request={Geo (x, y), dist, Hot}, and by Request={Geo (x, y), dist, Hot} is transmitted to LBSN database (2);

Hot spot region event computing module (33) second aspect is according to Request={Geo (x, y), dist, Hot} can search out (the x with Geo in LBSN database (2), y) hot spot region of coupling, is designated as inquiry return message Q _2-33;

Hot spot region event computing module (33) third aspect according to the frequency strategy POI-TP that registers under time window to described Q _2-33carry out computing, obtain region focus incident and rank ChecFreq, and described ChecFreq is fed back to user (1).

2. the hot spot region incident detection system based on geographical labels that is applied to LBSN network according to claim 1, is characterized in that: described geographical labels cluster strategy POI-CP has the following step:

Extraction belongs to the same area piece regX _yin the step of geographical labels POI;

Calculating belongs to the same area piece regX _yin the position number of geographical labels POI step;

Calculate the place-centric point of geographical labels POI and geographical labels position between maximum linear distance then described in judgement with zone radius threshold value r _{threshold value}size, if by r _{threshold value}assignment is to the region unit distance correlation radius in affiliated area if choose maximum linear distance as the region unit distance correlation radius in affiliated area and then by the distance correlation radius of a geographical labels distance correlation radius with b geographical labels sum is than any two geographical labels a in upper geographical labels POI, the central point distance between b obtain distance correlation $H\_\mathrm{re}{l}_{a-b}^{\mathrm{reg}{X}_y}=\frac{r{D}_a^{\mathrm{reg}{X}_y}+r{D}_b^{\mathrm{reg}{X}_y}}{\mathrm{CL}{D}_{a-b}^{\mathrm{reg}{X}_y}}$ Step;

Calculate any two geographical labels a in geographical labels POI, between b semantic dependency $S\_\mathrm{re}{l}_{a-b}^{Q_{31-32}}=1-\frac{E_{a-b}^{Q_{31-32}}}{\max \left(L_{a-b}^{Q_{31-32}}\right)}$ Step;

According to described with described with distance correlation threshold value rel _distance, semantic dependency threshold value rel _semanticcontrast, and merge region unit regX according to comparing result _ystep;

If and time, by the position of registering of b geographical labels merge to the position of registering of a geographical labels

If or time, the position of registering of b geographical labels not with the position of registering of a geographical labels carry out region unit merging.

3. the hot spot region incident detection system based on geographical labels that is applied to LBSN network according to claim 1, is characterized in that: the concrete mode of the frequency strategy POI-TP that registers under described time window is: hot spot region event computing module (33) can calculate in real time user (1) and ask any region regX _yfocus incident, register and record Q when ask separately in region to contain geography by the end of the history of current time t to focus region clustering database (2) _2-33, event description is $\mathrm{Trend}=\frac{{\mathrm{ChecFreq}}^t-{\mathrm{ChecFreq}}^{t-1}}{\mathrm{\ΔT}},$ Δ T represents time window, Δ T=|t-(t-1) |, t represents current time, t-1 represents previous moment, ChecFreq ^trepresent the frequency of registering of current time t, ChecFreq ^t-1represent the frequency of registering of previous moment t-1;

The active degree Rank of event is directly proportional with the event of continuing of registering to the frequency of registering of Δ T, that is:

$\mathrm{Rank}=\underset{i=1}{\overset{\mathrm{reg}{X}_y}{\mathrm{\Σ}}}\frac{{\mathrm{ChecFreq}}_i^t}{{\mathrm{\Ω}}_i}\×\sqrt{1+\frac{\underset{j\∈\mathrm{reg}{X}_y}{\max }\left({\mathrm{CU}}^j\right)}{\mathrm{reg}{X}_y}};$

Wherein be illustrated in the frequency of registering in time window Δ T, regX _yrepresent any region, the element of suing for peace, i represents summing target, Ω _ibe illustrated in time window Δ T user (1) the scope inner region of the asking total quantity of registering, represent the time window number in current all hot spot regions maximal value, j represents the area identification number of time window number maximum.