TRAP@NCI

Crime Prediction: Integration of data mining techniques (clustering and classification) to enhance crime prevention through analysis of the relationships between type of crimes, locations, times and weather patterns.

Chikarakate, Tinashe (2014) Crime Prediction: Integration of data mining techniques (clustering and classification) to enhance crime prevention through analysis of the relationships between type of crimes, locations, times and weather patterns. Diploma thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Diploma)
Download (2MB) | Preview

Abstract

The goal of this dissertation is to use data mining techniques (clustering and classification) to analyse the relationship of variables type of crime, occurrence time of crime, location of crime and weather pattern when crime happens. Being able to use the best data mining techniques to obtain the relationship between specific variables will help in coming up with a solution through crime prediction. Specific crime predications will be of greater importance for the world at large. The aim of this dissertation is to reveal the hidden characteristics and relations in crime by exploring the Seattle Police crime dataset and National climatic data center weather dataset and determining the model which can predict crime with outmost accuracy.

The general increase in crime in the world as a whole over the years inspired me to explore the data mining tools which are available to carry out crime analysis so as to come up with crime prediction model before unforeseen criminal activities happen.

For this analysis I used the k-means clustering algorithms to come with 5 clusters of crime. These clusters revealed that there are areas which require attention more than others due to the intensity of crimes. To further analyse these clusters I went on to apply classification algorithms to classify set of attributes to a particular class. I compared the algorithms Naïve Bayes and Decision trees (J48) I found out that decision trees are the best model with 95% accuracy correctly instances. For visualisation of the results from clustering I used Fusion google API for the crime heat maps.

For testing and evaluation of the models I used training and test datasets for validation purposes where there was need.

For future work I plan to use other algorithms other than the ones I have used and compare the results to see which one provides better results.

Item Type: Thesis (Diploma)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions: School of Computing > Higher Diploma in Science in Data Analytics
Depositing User: CAOIMHE NI MHAICIN
Date Deposited: 16 Dec 2014 17:25
Last Modified: 16 Dec 2014 17:25
URI: http://trap.ncirl.ie/id/eprint/1901

Actions (login required)

View Item View Item