A Geo-Statistical Approach for Crime hot spot Prediction

Abstract

Crime hot spot prediction is a challenging task in present time. Effective models are needed which are capable of dealing with large amount of crime dataset and prediction of future crime location. Spatio-temporal data mining are very much useful for dealing with the geographical crime data. In this paper sparse matrix analysis based spatial clustering technique for serial crime prediction model is used. Firstly, crime data are preprocessed through various distribution techniques and then sparse matrix analysis based spatial clustering technique are applied on a four years time series data from 2010 to 2014 for the major cities of India like Delhi, Mumbai, Kolkata and Chennai to find out the hotspot location for next year, after that three clustering techniques are used to grouping similar crime incident, at last cluster results obtained by original and proposed dataset are compared. The main objective of this research is applying crime prediction technique, forecast and detect the future crime location and its probability.