Crime prediction in Trinidad and Tobago using big data analyticsPredictive policing in developing countries

Crime undermines human and economic growth across all demographics. This is particularly true for developing countries. Thus, the reduction and prevention of crime has been a major focus for governments in Caribbean countries including Trinidad and Tobago (T &T). Big data analytics (BDA) has been extremely popular in exploring, identifying and predicting crime patterns. In this paper we use BDA techniques, such as exploratory data analysis (EDA), geocoding for hotspot mapping (GHM), kernel density estimation (KDE), and Twitter police advisement word-cloud (T-PAW) to analyse historical crime data and predict crime. We show each technique is individually robust providing valuable results. Our analysis showed breaking offences had the highest Prediction Accuracy Index (PAI) of 6.99 in 2020. We further demonstrated that crime data and Twitter data are both clustered in similar geographical areas confirming Twitter data is relevant in T &T crime analysis. Our ablation study shows adding Twitter data to the KDE technique resulted in a 9% improvement in accuracy. Authorities in developing countries may now consider using these techniques in reducing crime.