Data-Driven Prediction System of Dynamic People-Flow in Large Urban Network Using Cellular Probe Data

Box 1

Cellular raw data preprocessing algorithm.

Algorithm1: Redundant data processing (Data ordered by MSID, Timestamp)

t: timestamp in second; i = order number of raw data;

for all MSID u

If u.i.t = u.(i+1).t

Delete u.i;

end if

end for

If u.i.lat = u.(i+1).lat && u.i.lon = u.(i+1).lon && u.i.lat = u.(i-1).lat && u.i.lon = u.(I - 1).lon

Delete u.i;

end if

Algorithm2: Ping-Pong switching processing (Data ordered by MSID, Timestamp)

for MSID u

if |u.(i-1).t - u.(i+1).t| <T && u.(i-1).lat = u.(i+1).lat && u.(i-1).lon = u.(i+1).lon

Delete row: u.i;

end if

end for

Algorithm3: Drift switching processing (Data ordered by MSID, Timestamp)

for MSID u

if |u.(i-1).t - u.(i+1).t| <T &&  Δd/Δt > V && Δd > D

Delete row: u.i

end if

end for