We also realized that, in order to achieve high precision and recall
accuracy, only a subset of our initial set of classification features is
necessary. In addition to traditional features on average speed and
average acceleration, we identified for the first time the features on average bus closeness, average rail line closeness, and average
candidate bus closeness. Using only this subset of features, and
suppressing the other classification features that are not necessary,
the precision accuracy was still over 92.5%.