Table 14.

Classification results for real trajectories. The data set with $\sigma =0.38$ (no noise) and with the full sets of features was used for training. The numbers may not add up precisely to 100% due to rounding.

		Random Forest		Gradient Boosting
Features	Classified Mode	Receptor	G Protein	Receptor	G Protein
	Free diffusion	42%	40%	36%	35%
Set A	Subdiffusion	56%	54%	61%	58%
	Superdiffusion	1%	5%	1%	5%
	Free diffusion	51%	38%	44%	24%
Set B	Subdiffusion	44%	50%	37%	44%
	Superdiffusion	3%	10%	17%	30%
	Free diffusion	54%	51%	54%	51%
Set C	Subdiffusion	45%	47%	45%	46%
	Superdiffusion	0%	1%	0%	1%