Classification and Anomaly Detection in Traffic Patterns of New York City Taxis: A Case Study in Compound Analytics

Both graph-theoretical algorithms and machine learning techniques have been applied individually to solve a wide variety of transportation, network and routing problems. Many applications, however, are amenable to a combined approach. In this paper, we present a case study on the analysis of New York City taxi traffic using our compound analytics framework. Using incremental probabilistic learning, we infer a graph from each day's traffic. We then calculate a set of graph-theoretic metrics designed to capture a broad overview of the topology of the graphs. Using machine learning to reason over these metrics, we classify the graphs as coming from weekend or weekday traffic, and identify graphs affected by a range of disruptive events.