Table 2. Performance measures (AC, accuracy; FS, F1 score; KP, Kappa statistic) across class distributions for the Random Forest and HMM classifiers for dynamic data, forward roadway vs. instrument cluster, vs. left mirror, vs. center stack, vs. right mirror.

Forward vs.	Model	Original dataset			Balanced Dataset
		AC	FS	KP	AC	FS	KP
Instrument cluster	Random forest	0.59	0.33	0.08	0.56	0.48	0.11
Instrument cluster	Hidden Markov Model	0.66	0.32	0.12	0.66	0.61	0.33
Left mirror	Random forest	0.86	0.30	0.24	0.79	0.78	0.59
Left mirror	Hidden Markov Model	0.84	0.28	0.22	0.83	0.68	0.33
Center stack	Random forest	0.85	0.77	0.66	0.83	0.81	0.66
Center stack	Hidden Markov Model	0.83	0.72	0.60	0.85	0.83	0.69
Right mirror	Random forest	0.95	0.74	0.72	0.90	0.89	0.80
Right mirror	Hidden Markov Model	0.93	0.69	0.65	0.87	0.73	0.66