on on TimeTravlr TimeTravlr Simon Gélinas Simon Gélinas Because everyone loves getting stuck in traffic Because everyone loves getting stuck in traffic
Aug 07, 2015
ononTimeTravlrTimeTravlr
Simon GélinasSimon Gélinas
Because everyone loves getting stuck in trafficBecause everyone loves getting stuck in traffic
173 million taxi entries173 million taxi entries (~30gb) - Chris Whong FOIL (~30gb) - Chris Whong FOIL
Traffic data for ~1500 street segmentsTraffic data for ~1500 street segments - NYC open data - NYC open data
Raw data on Google BigQuery
Clean up
Extract patterns
Anomaly detection
Traffic Data: Traffic Data: Finding the underlying patterns Finding the underlying patterns
using singular value decomposition (SVD)using singular value decomposition (SVD)R
oad
ID #
Extremely strong within-week correlation!
Mon
Tu
e
W
ed
T
hu
F
ri
Sat
S
un
Mon - Tue - Wed - Thu - Fri - Sat - Sun
Reconstructed Geodata - TimeseriesReconstructed Geodata - Timeseries
Fast and accurate solution for websiteFast and accurate solution for website
Google map prediction
Scientific work published in: Nature (2013), Science (2014), Nature Materials (2014)
Simon GélinasSimon Gélinas
Traffic speed vs car densityTraffic speed vs car density
not very linear!not very linear!
Few cars → lots of cars
Fas
t
→
Slo
w
Traffic Data: Traffic Data: Position of Road ID # Position of Road ID #
Road ID #
MonTueWedThuFriSatSun