Using Bayesian online changepoint detection to reveal the impact of weather on individual-level cycling frequencies

Abstract

The impact of weather on cycling demand was usually investigated from a collective perspective. However, the impact on individual-level cycling behavioral changes over time was overlooked in the literature. This study addresses this gap using Bayesian online changepoint detection (BOCD) for behavioral change identification and discrete choice models for weather effect estimation. The proposed method was applied to reveal the impact of weather on cycling frequencies measured by weekly cycling days. We used a GPS dataset from Zurich comprising 520 cyclists with observation periods from 31 to 47 weeks. Results show relatively stable cycling frequencies in our sample, with an average of 2.3 changepoints detected per individual. Snow and precipitation are found to be the most significant attributes for driving a decrease in long-term cycling frequencies, while the effect of temperature remains inconclusive.