Smarter outlier detection and deeper understanding of large-scale taxi trip records: a case study of NYC

Abstract

Outlier detection in large-scale taxi trip records has imposed significant technical challenges due to huge data volumes and complex semantics. In this paper, we report our preliminary work on detecting outliers from 166 millions taxi trips in the New York City (NYC) in 2009 through efficient spatial analysis and network analysis using a NAVTEQ street network with half a million edges. As a byproduct of large-scale shortest path computation in outlier detection, betweenness centralities of street network edges are computed and mapped. The techniques can be used to help better understand the connection strengths among different parts of NYC using the large-scale taxi trip records.