Mobility pattern recognition method: Segmentation and geographic projection

Abstract

In this paper, a new spatial temporal pattern recognition technique specifically tailored to analyze mobility patterns is developed. The analysis is executed in the dimensions of time and space while taking into account the types of activities performed by population. The mobility patterns are transformed into sequences of strings through segmentation in the dimension of time, and a series of geographic projection operations in the dimension of space. The sequences are compared using Sequence Alignment Method (SAM) generating dissimilarity matrices and through data mining, clusters of mobility patterns along with their representative patterns are emerged. Using household travel survey data collected in New York City, we conduct an experiment to illustrate the application of the methodology and to demonstrate its capability in addressing questions regarding pattern disparity analysis and recognition on transportation data.