Spatio-temporal comparative analysis of scooter share in Washington D.C.

Abstract

Geospatial-temporal data for different e-scooter firms was collected and investigated for differences in e-scooter usage patterns among customers of the firms. Computational analysis using predictive algorithms and correlation analysis was done to find co-relationally important features for predicting the dependent variable. Data-preprocessing included computing trips from geospatial data and dividing the city into smaller clusters for analysis using geohashes. Hourly weather data was added to the geospatial temporal data to account for weather impact on the number of trips. The Spatio-temporal analysis shows a correlation between the percentage of scooters parked at a location and the success rate of the firm with the highest scooters getting the highest number of trips.