| At current times, China is experiencing a special period of transition. With the rapid development of urbanization and industrialization, various kinds of contradictions emerge continuously. Crimes, which jeopardize the lives, properties and social orders seriously, continue to rise. The pattern of urban crimes is not random, for it has close links with population, environment, economy, policy as well as social elements and presents some specific time and space characters. It is very important to get to know the temporal-spatial patterns of crimes because it plays a vital role in attacking crimes so as to improve people’s trust on public security sectors.Most of the current crime spatial analysis is qualitative and macroscopic. Although the time elements are also considered compared with early pure spatial perspective, it limits itself to the fixed time spans such as day, month, year etc. These methods cannot solve the problem of precise distribution of police forces and strong attacks against criminals.Scan statistics have several features:it could scan in both space and time at the same time; detection window could be dynamically changed; it could perform perspective analysis. Due to these facts, it has great potential in crime research field. Some researches show that scan statistics is better than other methods in detecting hotspots. Considering the fact that this method is to be designed for the detection of outbreak of diseases, it has limitations in crime field. This dissertation has optimized this method for crime analysis according to the characteristics of crimes. The main work is as follows:(1) Put forward improvement and optimization method to permutation statistics according to crime characteristics1) Put forward hotspot detection method to cases with line characteristicsBased on liner references, this dissertation puts forward the method of continuous grids to do scan statistics for those research objects with line characteristics. The first step is to process and transfer the crime data with coordinates by way of liner reference. Then perform scan statistics to2-demonsional cases with line characteristics, referenced by1-demensional scan statistics, on spatial or temporal-spatial analysis.2) Put forward abnormal period scan statistics according to seasonal characteristics of crime cases.According to seasonal characteristics of crime cases, this dissertation has smoothed expected value by way of Moving Index Weighted Average, which could reserve long term development trend so as to detect abnormal fluctuation. 3) Optimize the area-based scan algorithm by hexagon gridsBy transforming every single case to hexagon cell data, the new method could not only reduce the data amount to be analyzed but also model the hexagon grid by utilizing its spatial relations, which could change the retrieval method of the scan window. So, the speed of scan statistics has been greatly accelerated.4) Further process to scan statistics results based on GISBy further process to scan statistics results to detect the different types of hotspots namely, continuous, intermittent, emerging and disappearing.(2) Develop crime temporal spatial analysis software based on GISBased on Microsoft C#.NET platform and ESRI ArcGIS Engine API, the software is developed to perform crime temporal spatial scan statistics. By utilizing spatial topo-relations and spatial analysis functionalities of GIS, this software could not only improve the efficiency of scan statistics, but also realize the visualization of the results, which is helpful to detection of the police.(3) Research on cases of burglary and robbery-on-the-street by this softwareThis dissertation has done research on detecting temporal, spatial and temporal-spatial hotspots by means of software, choosing cases of burglary with area characteristics and cases of robbery-on-the-street with line characteristics within the period of2006to2010in Shanghai. And the conclusions are as follows:1)By analyzing the hotspots of burglary cases in time, this dissertation discovers that most of the occurance time is within November, December and January. Temporal hotspot within the short time span may happen in other months, such as Feb,6th2008(the Spring Festival Eve) and April,1st2009(April Fools’ Day). By analyzing the abnormal time span of burglary cases, this dissertation discovers that there are no anomalous time point with high significance. The lasting period is short generally,1to3days. The occurance time is in disperse.2)By analyzing the hotspots of burglary cases in space, this dissertation discovers that most of the occurance location covers the whole central part of the municipality, which shows that the the whole central part of the municipality is a big hotspot. While, the high risk hotspots are disperse in the perspective of the potential victims, for apart from the central part, high risk hotspots occur in subsurbs like chongming district, fengxian district and qingpu district. This dissertation also analyzes the the central part of the municipality in relatively small scale and finds that the distribution of hotspots are disperse. This dissertation also finds that small scale space analysis could detect the hotspots more precisely. While the large sacle may omit small area with high occurace rate. In addition, by post-process, this dissertation also finds the different types of hotspots like continuity, intermittence etc. 3)By analyzing the hotspots of burglary cases both in time and space, this dissertation discovers that the distribution of temporal-spatial hotspots is different from that based on pure spatial perspective. It is disperse both in space and time, for the occurrence time covers months from January to November continously. So, it has a large span both in time and space, which can not be detected neither in time nor in space analysis.4)By analyzing the hotspots of robbery-on-the-street cases in time, this dissertation discovers that most of the occurance time is at July and August. It detects the time span that cross the diffenent months by exceeding the limits of time span divided by traditional method. So it offers the more precisis results and it is evidenced by the detecting result, which is at middle and last part of August and the last part of July. Apart from that, the amouaus analysis basically has abandanded the seasonal element. And the results are not continous and it jumps in the time line. The results could help the police pay attention to other elments except for seasonal reason that lead to the happening of the hotspots.5)By analyzing the hotspots of robbery-on-the-street cases in space, this dissertation discovers that most of the occurance road segments, such as costal areas of BaoShan district, cross-border of Baoshan and ZhaBei, especially the Mudanjiang Road, Shuichan Road, Hutai Road, Changzhong Road. The occurance within these segments is lasting. The occurance of the cases at West Chanjiang Road is intermittent, and other places happens occasionaly.6)By analyzing the hotspots of robbery-on-the-street cases both in time and space, this dissertation also discovers that the distribution of temporal-spatial perspective is different from that of pure spatial or temporal perspective. It is disperse both in space and time, which shows that temporal-spatial analysis could finds hotspots happening at small area and within short period of time.7)This dissertation also analyzes the early warning time and its temporal-spatial evolvement pattern by utilizing historical data, which could be regarded as a reference for the early warning of crime clusters. For example, in time, there is a small probability to predict the next outbreak of the next day if there is only1outbreaking day as a reference(7/15); while the probability is bigger if there are2continous outbreaking days as a reference(7/9); As for the temporal-spatial early warning, although there are high probabilityof wrong report, the real outbreak basically could not be missed.All in all, Scan statestics analysis could show the patterns of the two types of crimes in time, space as well as temporal-spatial perspective, which offers the foundation for decision-making for the precise distribution of police forces and the crossponding prevension and control measures and also the refercence for early warning of crime clusters. |
PDF Full Text Request