Crime Prediction Using Geospatial Intelligence System for Crime Preventing

Crime prediction relies on significant amounts of various data sources and is analysed through mathematical models, predictive analytics techniques, and machine learning algorithms to identify patterns of crime. To date, there has been little research conducted to examine and extend which repeat and near-repeat victimisation within crime hotspots can be collocated for crime prediction (Chainey et al., 2018). Therefore, this paper aims to identify crime patterns using crime prediction with repeat and near-repeat analysis. This study employed the GIS tool, namely, repeat and near-repeat analysis as the primary methods. Historical crime: all types of data used in the years 2015 and 2016 are analysed. The area of study is Petaling Jaya, Selangor. By using repeat and near-repeat analysis, the results reveal that there is a significant (p=0.01) and a meaningful near-repeat victimisation pattern were found in the study area. The most over-represented space-time range that is significant is the zone from 1 to 100 metres and from 0 to 7 days from the initial incident. The likelihood of another crime incident is approximately 22% higher than if there is no discernible pattern. The most over-represented repeat victimisation range that is significant is the zone from 0 to 7 days after an initial incident. The likelihood of another incident is approximately 78 percent higher than if there were no repeat victimisation patterns. This result also shows there are five local hotspots as prediction zones in the study area. The importance of the study is that it provides useful information for assisting law enforcement in improving crime prevention strategies using geospatial technologies.