According to Figure 11, when the top five features are selected and
the other three classification features are pruned, the precision
accuracy hardly changes (see Figure 9 as well). Observe that
Random Forest (RF) classification model is still the dominating
classification model in Figure 11, with a precision and recall
accuracy of 92.8% and 92.9% respectively. For RF, only a 0.9 %
reduction in recall and precision accuracy is noticed when the five
top ranked features are considered, as opposed to considering all
eight classification features. This indicates that the top five
classification features are enough to detect transportation mode in
the proposed work.