Ant Colony Optimization-based Spatial Scan Statistic for Detecting Statistically Significant Spatial Communities in Vehicle Movements
2019-09-16T00:39:07Z (GMT) by
In the era of big data, discovery of spatial communities in vehicle movements plays a key role in understanding the urban structures and functions. While a number of community detection methods can be used to detect spatial communities in vehicle movements, these methods are usually designed without considering the network-constraint of vehicles and testing the significance of spatial communities. On that account, this study develops an ant colony optimization-based spatial scan statistic to detect statistically significant spatial communities in vehicle movements on urban road network. Road segments are used as basic units to represent the moving paths of vehicles. The spatial scan statistic is generalized to weighted spatially embedded graph to provide quantitative assessment for spatial communities, and the generalized spatial scan statistic and ant colony optimization are combined to detect arbitrarily shaped spatial communities. A Monte Carlo simulation method is developed to estimate the statistical significance of each detected spatial community. The effectiveness of the proposed method is evaluated by using both simulated and taxi GPS trajectory data sets.