Core Concepts
A novel sensor-based simulation method using the Discrete Empirical Interpolation Method (DEIM) can effectively detect spatiotemporal events in urban areas by identifying key locations and simulating uneventful scenarios.
Abstract
This paper presents a new method based on the Discrete Empirical Interpolation Method (DEIM) for event detection using point-based human mobility data. The key highlights are:
The method can first identify the dominant locations in the study area and then simulate an uneventful scenario based only on the limited observation data from the previously identified key locations.
Spatiotemporal events are detected by comparing the discrepancy between the observed actual data and the simulated uneventful scenario. Since this method is based on an unsupervised method, it does not require any prior knowledge of the study area or the time window of interest.
The method requires a lower data preparation effort as the simulation process is based on each discrete temporal unit and does not require chronologically ordered data.
The method is applied to analyze taxi trip records in New York City in 2009 and 2012. It can effectively detect major events such as Thanksgiving, Hurricane Sandy, St. Patrick's Day, and New Year's Eve, and identify the spatial distribution of the events.
Future research can explore different methods to determine the sensor locations, utilize the method with other data types, and experiment with different spatial and temporal unit combinations.
Stats
The taxi trip records in New York City in 2009 contained more than 143 million trips.
The study area was divided into 87,838 cells with a cell size of 110 m by 80 m.
Quotes
"This method can first identify the dominant locations in the study area and then simulate an uneventful scenario based only on the limited observation data from the previously identified key locations."
"Spatiotemporal events are detected by comparing the discrepancy between the observed actual data and the simulated uneventful scenario."