This classification feature relates the traveler’s GPS location with
spatial data representing the rail network. This feature may be
useful to detect if a person is travelling on an aboveground train.
For underground trains (subways), since GPS does not work well
underground, this feature may not be effective. We do not consider
subways in this work. The Euclidian distance dirail between a
person’s mobile device and the closest rail line is computed for
each GPS sensor report pi. We then calculate the classification
feature average rail location closeness (ARLC) as follows. Let {p1,p2, p3, p4…pn} be a finite the set of GPS reports submitted within a
time window.