Abstract:
Accurate spatio-temporal violent crime hotspot prediction is a difficult and challenging task at this present time. Large amount of violent crime dataset are usually required for predicting future occurrence of violent crime in terms of location and time. Various data mining techniques have been applied in the previous studies on violent crime prediction with accuracy and other results that needed to be improved upon. In this paper, Support Vector Machine based spatial clustering technique for violent crime prediction was used. Firstly, historical violent crime dataset between 2014 and 2019 Lagos, Nigeria were collected and preprocessed through Principal Component Analysis, and then Support Vector Machine model built using IBM Watson Studio was applied on the six different violent crime dataset collected to determine violent crime hotspot locations for next day in Lagos Nigeria. The model was evaluated using real-life dataset of six violent crime types (murder, arm robbery, kidnapping, rape, non-negligent assault and man slaughter) dataset using confusion matrix. The results obtained found to return an accuracy of 82.12 percent which is good to be relied on for violent crime prediction. Based on this result, the model could be used by the Police authority to develop new crime control strategies and plan towards mitigating crime rate in the country.
Page(s):
3187-3196
DOI:
DOI not available
Published:
Journal: Journal of Theoretical and Applied Information Technology, Volume: 98, Issue: 16, Year: 2020