Using machine learning models to predict the willingness to carry lightweight goods by bike and kick-scooter

The social transformation caused by the COVID-19 pandemic can help cities become healthier and more sustainable, with more space for active modes of transportation. This research addresses people's willingness to go shopping by bike or kick-scooter and to transport lightweight goods in cities with low maturity for cycling and scooting. Data collection was based on a survey, applied in the two largest cities of Brazil (São Paulo and Rio de Janeiro) and Portugal (Lisbon and Porto). The dataset was processed considering only two categories of respondents (i.e., potential users and regular users) and then four machine learning models (K-Nearest Neighbor, Support Vector Machine, Decision Tree, and Random Forest) were applied to predict shopping by bike or kick-scooter. In terms of all performance measures, the Support Vector Machine model was the optimum. The results indicate that people are willing to transport lightweight goods by bike or kick-scooter, as long as the infrastructure is safe and comfortable. This research contributes to understanding mobility behavior changes and identifying barriers to going shopping by bike or kick-scooter. It also presents some policy recommendations for improving cycling and scooting use for shopping, which public authorities can carry out.