| As informatization plays a more and more important role in the whole era,it is summarized in more fields,and police management crime and public security also need its penetration.The amount of measurement data in various fields of science has increased rapidly over the past few decades.This allows us to discover relationships that should exist between the measured variables.Detailed spatial data can be easily obtained through the Internet and a user-friendly geographic information system that manages network data,thus solving the previous technical difficulties to a great extent.In order to focus the provided exploratory data analysis on the relationships that most favor the desired results,an automated process is necessary,thus introducing the Kernel Density Crime Spatiotemporal Algorithm.DBSCAN algorithm is one of the most classical density clustering algorithms,the input data set D,radius Eps,density threshold Min Pts can be in the spatial database with noise to form arbitrary shape of the cluster.However,the selection of the parameters Min Pts and Eps,and the determination of the number of clusters greatly affect the quality of clustering.So we can use the adaptive method to determine the parameter size in order to better clustering.The kernel density estimation graph is used to represent the data distribution in the continuous data segment.The smoothing peak function is used to fit the known data points to simulate the real probability distribution curve.It is not affected by the number of packets and can define the distribution shape better.In this paper,kernel density estimation is used to calculate the density,so that the global threshold can be determined quickly and the data set can be identified effectively.In DBSCAN,the nearest neighbor method is used to measure the sample distance by some distance measure,such as Euclidean distance.However,for the road problem,such a calculation method is not desirable.This thesis explores the method of road network kernel density estimation,changes the previous distance calculation method to road network distance,and calculates the suitable value range of the parameter Eps of DBSCAN by combining with the kernel density,that is,combining the road network kernel density and DBSCAN to cluster traffic and crime and other road problems,namely the ODMK-DBSCAN algorithm.In the traditional model,it is generally believed that the plane space is isotropic,but in dealing with some problems,such as life-path-related,it is impossible to calculate the probability density in a circle.Therefore,point events based on network constraints are adopted to analyze problems such as roads.Network kernel density estimation decomposes the trend of the road into equal linear cells.The points to be estimated are scattered along the network direction,and the total number of them in the unit direction is calculated.The probability density and the event intensity in the network are judged by the points to be evaluated,where the bandwidth is determined by the network path in the shortest cell.The network kernel density estimation is used to draw the K-distance curve to determine the optimal range of Eps,and the contour coefficient combining the cohesion and the separation factor is used to obtain the most suitable parameter value by evaluating the clustering quality.In order to ensure the feasibility of the algorithm,three indexes,accuracy rate,adjustment of Rand Coefficient and mutual information score,are used to evaluate the algorithm by comparing the experimental and real data set indexes in different dimensions,the superiority of checking the parameter of density estimation.Secondly,the global Molain index and the local spatial auto-correlation Molain index are calculated to test the distribution pattern of the spatial pattern of the crime rate in San Francisco,finally,the prediction of the crime rate of Tyson Polygon based on the spatial relation model is carried out,and the high-value clustering features are compared with ODMK-DBSCAN to verify the accuracy and validity of the algorithm. |
