According to Figure 10, ABC is a more effective classification
feature than CBC for transportation mode detection. In order to
quantify the contribution of CBC to bus mode detection accuracy
as a classification feature, we present bus mode accuracy results
using the Random Forest Model, when the CBC feature is
suppressed. The four high order features (average speed, average
acceleration, average rail closeness, and average bus closeness) of
Figure 10 are used in Figure 14. From Figure 14, we observe that
the precision accuracy of buses decreases to 85.1 % (Figure 14)
from 89.7% (Figure 11), when only the top four features are used.
This indicates that even though CBC may be more time consuming
to compute, and use more memory than ABC, it is a worthwhile
feature for buses. On the other hand, if speed and memory is
critical; CBC can be suppressed, and bus mode detection will
remain over 85% accurate