Online Program

Spatial-Temporal Change Detection Using Elastic Functional Data Analysis (309688)

Trisalyn Nelson, University of California Santa Barbara 
*Avipsa Roy, Arizona State University 
Pavan Turaga, Arizona State University 

Keywords: Strava, Elastic Functions, SRVF, K-Means, Change Detection

Active transportation (e.g. biking and walking) is a new paradigm in building healthier and safer cities. Local authorities can create opportunities for people to exercise for recreation as well as to incorporate physical activity into their daily routines. Monitoring change in active modes of transportation can help authorities to make improved decisions and to promote healthier living by evaluating trends in mobility patterns over time that highlight the need for better infrastructure. In this study, we focus primarily on monitoring change in bicycling ridership from big spatial-temporal data. With the emergence of crowdsourced fitness apps like Strava, a novel source of high-frequency bicycling data has become available to researchers. However, monitoring change remains a challenging problem due to sensitivity of models to varying sampling rates and temporal misalignment of data. Our goal is to present a new approach to detect changes in bicycling ridership patterns based on a functional data analysis (FDA) framework that enables analyzing big crowdsourced ridership data captured while addressing non-elastic rate variations in ridership. Using data from 177,137 street segments with 28.6 million instances of activities from the Strava fitness app recorded every minute, we detect ridership changes in the Phoenix Metropolitan area for the years 2017 and 2018. We quantify the change in ridership at the street-segment level for every hour of the day and generate maps to visulaize the spatial variation of change across the entire Phoenix metropolitan area. Our results indicate that 11.8\% of street segments show high change, 18.34\% segments show medium change, and 39.6\% show low change in ridership at the hourly scale. Most of these segments lie in the southern and central parts of Phoenix, which also exhibit a high density of existing or newly installed bicycling infrastructure, and experienced an increase of more than 50\% Strava riders annually from 2017.