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.
Citation
Masters of science in computer sciencePublisher
University of Nairobi School of Computing and Informatics