Project Summary
The recent emergence of GPS-enabled smartphone applications (apps) has facilitated a vast increase in capturing human movement data that is changing the landscape of AT research. Crowdsourcing platforms that leverage GPS technology by means of fitness tracking apps (eg: Strava, Streetlight, etc.) have therefore become a tool of interest for collecting data on pedestrian volumes and/or bicycling ridership. Such high-resolution data may help planners and decision-makers improve bicycling facilities and safety, raise modal share and help realize the personal fitness and environmental benefits of AT using Artificial Intelligence (AI) technologies if used cautiously. These types of smart device captured data are emerging as an important data source for evidence-based city planning, decision-making, and more efficient provision of urban services. However, the bias in crowdsourced data collected through apps leads to a major hindrance in utilizing the data directly into AI-driven transportation planning research. Unless corrected at the early stages of data preprocessing and conveyed to consequent stages of the AI pipeline, the biased samples can eventually lead to inequitable infrastructure/policy decisions made by practitioners. My research aims to directly address the fundamental aspects related to sampling bias in crowdsourced data at all stages of the AI pipeline to ensure fair distribution of transportation infrastructure in the Southern California region by using bias-adjustment factors that predict accurate estimates of ridership/pedestrian volumes representative of all populations at the street segment across multiple cities. The bias-corrected counts will be used to generate street-segment level maps of pedestrian and bicyclist volumes across multiple cities in Orange County.
