Multiple Naïve Bayes Classifiers Ensemble for Traffic Incident Detection

Table 3

Experimental Results of NB, Five Rules and NBTree Based as Applied to the I-880 Dataset (The performances are presented in the form average ± variance).

Algorithm	DR	FAR	MTTD	CR	AUC	Kappa	MAE	RMSE	EC
Naïve Bayes classifier	0.8228±2.50𝐸−05	0.0398 ± 6.32E − 07	1.2991 ± 0.01727	0.9540 ± 6.43E − 07	0.8915 ± 6.50E − 06	0.59479 ± 2.63E − 05	0.01619 ± 3.11E − 07	0.17995 ± 9.47E − 06	0.99184 ± 8.03E − 08
Product rule ensemble	0.8404 ± 2.91E − 04	0.0380 ± 3.03E − 06	1.7615 ± 1.04744	0.9557 ± 2.22E − 06	0.9012 ± 7.78E − 05	0.61383 ± 2.28E − 04	0.01602 ± 5.33E − 07	0.17898 ± 1.60E − 05	0.99192 ± 1.38E − 07
Sum rule ensemble	0.8963 ± 5.47E − 06	0.0297 ± 5.87E − 07	1.6906 ± 0.89473	0.9636 ± 3.66E − 07	0.9333 ± 1.06E − 06	0.6932 ± 1.91E − 05	0.01599 ± 2.04E − 07	0.17881 ± 6.34E − 06	0.99194 ± 5.26E − 08
Max rule ensemble	0.8960 ± 5.58E − 06	0.0297 ± 1.62E − 07	1.6236 ± 0.84537	0.9636 ± 1.63E − 07	0.9331 ± 1.33E − 06	0.69268 ± 6.81E − 06	0.01606 ± 8.09E − 08	0.17924 ± 2.52E − 06	0.99190 ± 2.09E − 08
Min Rule ensemble	0.8962 ± 2.29E − 06	0.0304 ± 2.99E − 06	1.6193 ± 0.85452	0.9629 ± 3.13E − 06	0.9329 ± 2.53E− 06	0.68845 ± 1.28E − 04	0.01603 ± 1.06E − 07	0.17902 ± 3.28E − 06	0.99192 ± 2.73E − 08
MV rule ensemble	0.8193 ± 6.25E − 05	0.0410 ± 2.72E − 06	1.7802 ± 1.15158	0.9527 ± 2.02E − 06	0.8892 ± 1.29E − 05	0.58639 ± 5.16E − 05	0.01630 ± 5.09E − 07	0.18052 ± 1.52E − 05	0.99178 ± 1.32E − 07
NBTree classifier	0.8143 ± 0.00112	0.0089 ± 1.463E − 5	1.3622 ± 0.00798	0.9831 ± 1.474E − 5	0.9027 ± 2.7953E − 4	0.8052 ± 0.00147	0.02626 ± 6.3487E − 6	0.22894 ± 1.202E − 4	0.98669 ± 1.67E − 6