Combating Motor Vehicle theft using Decision Support Models: Nairobi Case Study

Abstract

Forecasting is an essential analytical tool in police planning and allocation of resources. Law enforcement agents have complex databases which has intelligence hidden as trends, patterns, dependencies, and relationships. Data Mining is the process of acquisition of this knowledge from databases inform of significant patterns and associations. This project focuses on machine learning tools and identifies Artificial Neural Network model that can be used for motor vehicle theft short term trend and patterns forecasting in Nairobi. The study involves several experiments using WEKA, Zaitun Time Series, Neuralab and Tiberius software to forecast motor vehicle theft. The data was prepared to forecast geographical location where theft was likely to occur. The forecasted results were trivial and hence disregarded. The second set of experiment involved forecasting motor vehicle theft counts using time series Neural Network with WEKA software. The results were successful and able to forecast the motor vehicle theft tred for the succeeding 2 to 3 months. These results were further used to extract rules from the trained network. These rules explained future trends in motor vehicle theft. The last study was to identify an open source time series neural solution that could be used to support decision making to combat motor vehicle theft. Three software solutions were studied and Zaituni Time series software performed extremely well. The motor vehicle theft trend forecasted in this project confirmed the research hypothesis that artificial intelligence can be used to forecast crime trends.