How can we Apply Crime Prediction Technics to Japan?

• In Event: Posters I

Abstract

This research tries to examine an application possibility of crime prediction technics in Japan where fewer crimes occur than in other countries and number of crime incidents has been decreasing year after year. Since there has been little research on them in Japan, we have reviewed empirical studies on crime forecasting methods which focused on when and where crime would occur. Furthermore, we tested and evaluated some methods by applying them to actual crime data in Japan. Tested methods include hot-spot mapping using kernel density estimation method, Prospective hot-spot mapping (Bowers et al., 2004), Self-exciting point process method (Mohler et al., 2011), Risk terrain modeling (Caplan et al., 2011) and other regression methods. In order to compare them, we assessed Predictive accuracy, Compactness, Dynamic variability (Adepeju et al., 2016) and an easiness of performing. Finally we discussed about which method is the most valid and has a practical superiority in Japan.

Authors

• Tomoya Ohyama, University of Tsukuba

• Mamoru Amemiya, University of Tsukuba