Crime analysis and forecasting using spatial data mining

Abstract

The importance of Data Mining in analysing crime never ceases to increase. Criminal activity leads to serious problems all around the world, thus it is vital to solve crimes as fast as possible and to allocate resources effectively. Therefore, Data Mining can greatly improve and facilitate crime analysis by processing and extracting useful information from large volumes of data. This project revolves around the deployment of a data mining cycle and algorithms which take spatial data and criminal record data from Chicago, IL, USA, and output knowledge and patterns that classify crimes and crime patterns over time and place. Data Mining algorithms most often search huge amounts of data and possible correlations and combinations between data items. A whole database does not fit in main memory, and access to data on disk is significantly slower. Thus, an ongoing research challenge is to create and improve the speed and scalability of data mining algorithms. For this reason, another aim of this project will be to study and identify downfalls and possible improvements that can be made to current data mining algorithms in order to make them more efficient. This dissertation provides an implementation of the most well-known data clustering techniques used in Data Mining, and an analysis of how these can be utilised to facilitate the critical task of crime solving.