CN104765772A - Modeling method based on time-space regional criminal characteristics - Google Patents

Abstract

The invention belongs to the technical field of geological information science and data excavation, and refers to time-space classification of criminal risk, and criminal risk characteristic analysis, and particularly relates to a modeling method based on time-space regional criminal characteristics. The modeling method is to classify a criminal time-space region into R region, K themes, T semantic time, and W criminal parameter index. The modeling method comprises the following steps: (1) setting a potential criminal characteristics model in a region; (2) classifying and modeling the time state of a criminal level; (3) generating and modeling the criminal level; (4) generating and modeling a criminal parameter index. The modeling method has accuracy and effectiveness in predicting related regional criminals, and has important functions in improving the prediction and analysis of the criminal level of a suspect, and assessing the regional criminal risk of the suspect.

Description

A kind of modeling method based on spatio-temporal region characteristics of crime

Technical field

The invention belongs to Geographical Information Sciences, data mining technology field, relate to the division of crime risk space-time, crime risk signature analysis, particularly relate to a kind of modeling method based on spatio-temporal region characteristics of crime.

Background technology

Social economic environment is the main cause affecting suspect's crime case space distribution, and different social economies and the interaction of criminal offence can show local characteristics of crime.Here; the specific crime chance that we claim region segmental society economic environment to build is region " potential characteristics of crime "; the success that daily routines theory proposes crime must meet three key elements simultaneously: potential criminal; suitable target; the disappearance that crime is taken precautions against; therefore also " potential characteristics of crime " can be regarded as the degree converged of crime three elements because people's daily routines determine, and the features of regional environment formed thus.

Crime is (working day in different semantic times, the Spring Festival etc.), due to reasons such as society's work and rest rule, social usage, weather and social activitieies, criminal motive, crime wish, the crime means etc. of crime one's share of expenses for a joint undertaking also can change, we are referred to as " the potential characteristics of crime " of semantic times, " the potential characteristics of crime " of region " potential characteristics of crime " and semantic times by region crime factor indirect expression out, suspect in the interregional distribution of the potential characteristics of crime of difference, often contain their position transfer mode and translate into because of.Therefore, this chapter object is exactly to describe the potential characteristics of crime with quantization areas, to support the research of subsequent sections characteristics of crime difference measure, suspect's zone-transfer pattern, to quantize and crime correlated characteristic, support for providing spatio-temporal prediction analysis based on characteristics of crime.

Summary of the invention

Object of the present invention is the problems referred to above solving prior art, and provide a kind of modeling method based on spatio-temporal region characteristics of crime, to achieve these goals, the technical solution used in the present invention is as follows:

Based on a modeling method for spatio-temporal region characteristics of crime, it is characterized in that: described modeling method crime spatio-temporal region is divided into R region, K theme, T semantic times, and W crime parameter index, comprises the following steps:

(1) the potential characteristics of crime model in region is set up, for describing and quantizing the ratio in potential characteristics of crime distribution situation analyzed area shared by every latency, thus the potential crime levels in this region of comprehensive assessment; Described characteristics of crime model expression is: R=(W, T, c, θ), wherein, W represents the property value of this region crime parameter indices, T represents each semantic times residing for this region, and c represents the crime levels of each semantic times, and θ represents the potential characteristics of crime in district;

(2) tense of crime levels divides modeling, for describing the crime distribution of region in different semantic times, thus can the potential feature of analyzed area more comprehensive and accurately, the potential characteristics of crime distribution under different semantic times u is expressed as: P (u|t)=Muti (z _t1, z _t2..., z _tk), wherein, z _tkfor the potential characteristics of crime extracted in K theme;

(3) crime levels generates modeling, and for describing the potential characteristics of crime of the potential characteristics of crime in region and each semantic times, carry out distribution assessment to the crime levels of zones of different, the distributed model of the individual region crime levels in zoning is: c _ij~ N (u _i ^tθ _j, σ _ij), wherein, c _ijfor the crime quantity of period i lower area j, u is the potential characteristics of crime distribution under i-th period, θ _jfor the potential characteristics of crime distribution of region j, σ _ijfor the medium factor in region;

(4) crime parameter index generates modeling, analyzing the index distribution that institute zoning affects potential characteristics of crime, then predicting the space-time characteristics of crime in this region for generating.

Preferably, the medium factor in described region is: σ _ij=ξ | L _ij(w _j1, w _j2..., w _j|W|)-c _ij|, wherein, | W| is the quantity of all economic environment pointer types, and ξ is constant, L _ij(.) is multiple linear regression model, for the crime parameter index of descriptive semantics time i and region j.

Preferably, the constant ξ span in the medium factor in described region is 0.05 ~ 0.2.

Preferably, described crime parameter index adopts multiple linear regression model L _ij(.) distribution carries out modeling, described multiple linear regression model L _ijthe distribution of (.) is as follows:

$L=-\frac{{\mathrm{\λ}}_u}{2}{\mathrm{\Σ}}_i{u}_i^T{u}_i-\frac{{\mathrm{\γ}}_v}{2}{\mathrm{\Σ}}_j{(v_j-{\mathrm{\θ}}_j)}^T(v_j-{\mathrm{\θ}}_j)+{\mathrm{\Σ}}_j{\mathrm{\Σ}}_n\log \left({\mathrm{\Σ}}_k{\mathrm{\θ}}_{\mathrm{jk}}{\mathrm{\β}}_{k,w_{\mathrm{jn}}}\right)-{\mathrm{\Σ}}_{i,j}\frac{{\mathrm{\σ}}_{\mathrm{ij}}}{2}{(c_{\mathrm{ij}}-u_i^T{v}_j)}^2,$

Wherein, β is the distribution of potential characteristics of crime-crime parameter index, and u is theme distribution under semantic times, ν is area topic distribution.

In sum, the present invention has following beneficial effect:

The present invention is by the modeling analysis of spatio-temporal region characteristics of crime, carry out space-time division to region to carry out predicting that characteristics of crime analysis provides important theory support, and predict accuracy and the validity of relevant range crime, there is important effect to raising forecast analysis suspect crime levels, assessment suspect region crime risk.

Accompanying drawing explanation

In order to be illustrated more clearly in example of the present invention or technical scheme of the prior art, introduce doing accompanying drawing required in embodiment or description of the prior art simply below, apparently, accompanying drawing in the following describes is only examples more of the present invention, to those skilled in the art, do not paying under creationary prerequisite, other accompanying drawing can also obtained according to these accompanying drawings.

Fig. 1 is the principle model of a kind of modeling method based on spatio-temporal region characteristics of crime of the present invention.

Fig. 2 is the region crime quantity figure of a kind of modeling method based on spatio-temporal region characteristics of crime of the present invention.

Fig. 3 is that the average crime levels of the region original topic areal distribution of a kind of modeling method based on spatio-temporal region characteristics of crime of the present invention affects Error Graph.

Fig. 4 is that the region potential characteristics of crime distribution of a kind of modeling method based on spatio-temporal region characteristics of crime of the present invention affects Error Graph to the average crime levels of the overall situation.

Embodiment

Below in conjunction with the accompanying drawing in example of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

As Fig. 1, a kind of modeling method based on spatio-temporal region characteristics of crime is that crime spatio-temporal region is divided into R region, K theme (potential characteristics of crime), T semantic times, and W crime parameter index, mainly comprises the following steps:

(1) the potential characteristics of crime model in region is set up, for describing and quantizing the ratio in potential characteristics of crime distribution situation analyzed area shared by every latency, thus the potential crime levels in this region of comprehensive assessment; Described characteristics of crime model expression is: R=(W, T, c, θ), wherein, W represents the property value of this region crime parameter indices, T represents each semantic times residing for this region, and c represents the crime levels of each semantic times, and θ represents the potential characteristics of crime in district;

(2) tense of crime levels divides modeling, for describing the crime distribution of region in different semantic times, thus can the potential feature of analyzed area more comprehensive and accurately, the potential characteristics of crime distribution under different semantic times u is expressed as: P (u|t)=Muti (z _t1, z _t2..., z _tk), wherein, z _tkfor the potential characteristics of crime extracted in K theme;

(3) crime levels generates modeling, and for describing the potential characteristics of crime of the potential characteristics of crime in region and each semantic times, carry out distribution assessment to the crime levels of zones of different, the distributed model of the individual region crime levels in zoning is: c _ij~ N (u _i ^tθ _j, σ _ij), wherein, c _ijfor the crime quantity of period i lower area j, u is the potential characteristics of crime distribution under i-th period, θ _jfor the potential characteristics of crime distribution of region j, σ _ijfor the medium factor in region; Wherein:

The medium factor in described region is: σ _ij=ξ | L _ij(w _j1, w _j2..., w _j|W|)-c _ij|, wherein, | W| is the quantity of all economic environment pointer types, and ξ is the medium factor constant in region, and span is 0.05 ~ 0.2, L _ij(.) is multiple linear regression model, for the crime parameter index of descriptive semantics time i and region j; In the present invention, the medium factor is the correction to the potential characteristics of crime in all regions, and passes through variances sigma _ijlookup protocol, the potential characteristics of crime of tense is also revised, more accurately to express the property value of each region crime parameter indices under each semantic times.

In the present invention, described crime parameter index adopts multiple linear regression model L _ij(.) distribution carries out modeling, described multiple linear regression model L _ijthe distribution of (.) is as follows:

$L=-\frac{{\mathrm{\λ}}_u}{2}{\mathrm{\Σ}}_i{u}_i^T{u}_i-\frac{{\mathrm{\γ}}_v}{2}{\mathrm{\Σ}}_j{(v_j-{\mathrm{\θ}}_j)}^T(v_j-{\mathrm{\θ}}_j)+{\mathrm{\Σ}}_j{\mathrm{\Σ}}_n\log \left({\mathrm{\Σ}}_k{\mathrm{\θ}}_{\mathrm{jk}}{\mathrm{\β}}_{k,w_{\mathrm{jn}}}\right)-{\mathrm{\Σ}}_{i,j}\frac{{\mathrm{\σ}}_{\mathrm{ij}}}{2}{(c_{\mathrm{ij}}-u_i^T{v}_j)}^2$

Wherein, β is the distribution of potential characteristics of crime-crime parameter index, and u is theme distribution under semantic times, ν is area topic distribution;

(4) crime parameter index generates modeling, analyzing the index distribution that institute zoning affects potential characteristics of crime, then predicting the space-time characteristics of crime in this region for generating.

Embodiment 1:

In the present embodiment, choose 3601 regions (community), what wherein region crime dramas was maximum is 145, minimum is 0,2.9, average each community, has burglary crime dramas and records 1945 regions district, in that region average out to 5.4, for crime dramas as shown in Figure 2, can find to only have the region of 1 crime dramas to account for the overwhelming majority.

Composition graphs 2, λ _vdistribute as region original topic, medium factor sigma _ij, its value is larger, and the deviation that area topic distribution and original topic distribute is less.Here, model parameter K=20, λ _u=0.01, ξ=0.2.λ _vvalue arrange respectively 50,100,150,200,250,300,350,400,450,500}, predicts its impact on crime quantity, and carries out the average crime levels error E rr of the judge overall situation:

$\mathrm{Err}=\frac{{\mathrm{\Σ}}_{i\∈T}{\mathrm{\Σ}}_{j\∈R}|c_{\mathrm{ij}}^{\mathrm{real}}-c_{\mathrm{ij}}^{\mathrm{test}}|}{\left|T\right|\left|R\right|}$

Wherein, R is the quantity in region, for the actual crime quantity of i semantic times lower area j, for the crime levels that this semantic times and this region drag are estimated, as shown in Figure 3, at λ _vtime less, medium factor sigma _ijimpact larger; At λ _vbecome greatly, medium factor sigma _ijimpact diminish, therefore by assessing preferably crime levels region drag estimation prediction, obtain theme distribution to portray crime crime levels, according to medium factor sigma _ijcrime levels is assessed to the correction of original topic distribution.

Embodiment 2:

In this enforcement, K the theme (potential characteristics of crime) divided, to also having material impact because whether quantum count K accurately expresses the potential characteristics of crime in region to it in potential characteristics of crime, K arranges too much or too small, is all unfavorable for accurately portraying features of regional environment, in the present embodiment, K value is set { 10,20,30,40,50,60,70,80,90, change between 100}, the overall average error Err of the crime levels that observation model obtains and true horizon, decide final K value.Wherein, λ _u=0.01, λ _v=200, ξ=0.2, as shown in Figure 4, along with the increase of K value, more accurate to the description of provincial characteristics, therefore Err reduces gradually, when K value reaches 30 time, Err=2.75 is overall minimum, after this, the increase of K causes Err fluctuation to be risen, when K is 80 time, Err is maximum, reach 5.94, this represents that theme number is too much, is unfavorable for improving model to the forecast analysis of crime levels and assessment.

The foregoing is only the preferred embodiment of invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (4)

1. based on a modeling method for spatio-temporal region characteristics of crime, it is characterized in that: described modeling method crime spatio-temporal region is divided into R region, K theme, T semantic times, and W crime parameter index, comprises the following steps:

(1) the potential characteristics of crime model in region is set up, for describing and quantizing the ratio in potential characteristics of crime distribution situation analyzed area shared by every latency, thus the potential crime levels in this region of comprehensive assessment; Described characteristics of crime model expression is: R=(W, T, c, θ), wherein, W represents the property value of this region crime parameter indices, T represents each semantic times residing for this region, and c represents the crime levels of each semantic times, and θ represents the potential characteristics of crime in district;

(2) tense of crime levels divides modeling, for describing the crime distribution of region in different semantic times, thus can the potential feature of analyzed area more comprehensive and accurately, the potential characteristics of crime distribution under different semantic times u is expressed as: P (u|t)=Muti (z _t1, z _t2..., z _tk), wherein, z _tkfor the potential characteristics of crime extracted in K theme;

(3) crime levels generates modeling, and for describing the potential characteristics of crime of the potential characteristics of crime in region and each semantic times, carry out distribution assessment to the crime levels of zones of different, the distributed model of the individual region crime levels in zoning is: c _ij~ N (u _i ^tθ _j, σ _ij), wherein, c _ijfor the crime quantity of period i lower area j, u is the potential characteristics of crime distribution under i-th period, θ _jfor the potential characteristics of crime distribution of region j, σ _ijfor the medium factor in region;

(4) crime parameter index generates modeling, analyzing the index distribution that institute zoning affects potential characteristics of crime, then predicting the space-time characteristics of crime in this region for generating.

2. a kind of modeling method based on spatio-temporal region characteristics of crime according to claim 1, is characterized in that: the medium factor in described region is: σ _ij=ξ | L _ij(w _j1, w _j2..., w _j|W|)-c _ij|, wherein, ξ is constant, L _ij(.) is multiple linear regression model, for the crime parameter index of descriptive semantics time i and region j.

3. a kind of modeling method based on spatio-temporal region characteristics of crime according to claim 2, is characterized in that: the constant ξ span in the medium factor in described region is 0.05 ~ 0.2.

4. a kind of modeling method based on spatio-temporal region characteristics of crime according to claim 1 and 2, is characterized in that: described crime parameter index adopts multiple linear regression model L _ij(.) distribution carries out modeling, described multiple linear regression model L _ijthe distribution of (.) is as follows:

$L=-\frac{{\mathrm{\λ}}_u}{2}{\mathrm{\Σ}}_i{u}_i^T{u}_i-\frac{{\mathrm{\γ}}_v}{2}{\mathrm{\Σ}}_j{(v_j-{\mathrm{\θ}}_j)}^T(v_j-{\mathrm{\θ}}_j)+{\mathrm{\Σ}}_j{\mathrm{\Σ}}_n\log \left({\mathrm{\Σ}}_k{\mathrm{\θ}}_{\mathrm{jk}}{\mathrm{\β}}_{k,w_{\mathrm{jn}}}\right)-{\mathrm{\Σ}}_{i,j}\frac{{\mathrm{\σ}}_{\mathrm{ij}}}{2}{(c_{\mathrm{ij}}-u_i^T{v}_j)}^2,$

Wherein, β is the distribution of potential characteristics of crime-crime parameter index, and u is theme distribution under semantic times, ν is area topic distribution.