Mining large-scale human mobility data for long-term crime prediction

EPJ Data Science

Table 3 Temporal out-of-sample results of the regressors using different subsets of the features: trained on 2014 and tested on 2015

	Census		Census + POI		Human Dynamics		Census + Human Dynamics
	MSE	\(R^{2}\)	MSE	\(R^{2}\)	MSE	\(R^{2}\)	MSE	\(R^{2}\)
Total incidents
Random Forest	0.11	0.82	0.07	0.88	0.09	0.84	0.07	0.88
Extra-Tree	0.11	0.82	0.07	0.89	0.08	0.87	0.07	0.89
Gradient Boosting	0.22	0.64	0.09	0.85	0.12	0.80	0.08	0.87
Grand larcenies
Random Forest	0.19	0.73	0.14	0.81	0.14	0.81	0.13	0.82
Extra-Tree	0.21	0.71	0.14	0.81	0.14	0.80	0.14	0.80
Gradient Boosting	0.28	0.61	0.17	0.77	0.16	0.78	0.15	0.79
Robberies
Random Forest	0.27	0.71	0.24	0.75	0.28	0.70	0.23	0.75
Extra-Tree	0.26	0.72	0.23	0.75	0.27	0.70	0.27	0.71
Gradient Boosting	0.38	0.59	0.29	0.69	0.32	0.66	0.28	0.70
Burglaries
Random Forest	0.25	0.47	0.24	0.50	0.25	0.47	0.24	0.49
Extra-Tree	0.25	0.47	0.24	0.50	0.33	0.32	0.32	0.34
Gradient Boosting	0.30	0.38	0.25	0.47	0.27	0.42	0.23	0.51
Assaults
Random Forest	0.24	0.75	0.22	0.77	0.28	0.72	0.22	0.77
Extra-Tree	0.24	0.76	0.22	0.78	0.28	0.71	0.27	0.73
Gradient Boosting	0.34	0.65	0.29	0.71	0.46	0.53	0.24	0.76
Vehicle larcenies
Random Forest	0.31	0.31	0.29	0.34	0.31	0.31	0.30	0.34
Extra-Tree	0.33	0.27	0.29	0.36	0.37	0.16	0.38	0.15
Gradient Boosting	0.32	0.28	0.31	0.30	0.34	0.23	0.30	0.33