The application of SOM and K-means algorithms in public security performance analysis and forecasting

Abstract

As cluster analysis is the most commonly used Data Mining method in performance analysis and forecasting, this paper establishes an auxiliary model framework for the security system, using clustering techniques to analyze data. We compared the analysis results of SOM and k-means algorithms, testing for their strengths and weaknesses, combing these two algorithms, and finally proposing the two-stage clustering algorithm based on SOM and k-means algorithms. In this thesis, by clustering and analyzing the same or similar cases, combining with the recommendations of experts, and determining main factors affecting the performance among different regions, we provide references not only for the investigators who will analyze the cases, but also for the public security agencies in the performance prediction, police force forecast and later development of the public security warning system. © 2012 Springer-Verlag Berlin Heidelberg.