Analysis of the Demand for Bike Sharing in Oslo with Machine Learning

Abstract

This paper aims to create accurate predictive models for the bike-sharing system operated by Oslo City Bikes. The three different machine learning methods are used to predict user demand within a specified area of Oslo. Furthermore, the paper intends to discover which factors that most influence bike-sharing usage preto mid-pandemic. Different factors of bike-sharing systems will be evaluated to create a reliable model. Recommendations for further research topics, as well as possible business implementations for the model will be explained. The machine learning method with the best performance was GRU with an MAE score of 14.30, RMSE of 20.80 and R����� of 0.77. Multiple COVID-19 features indicating varying intensities of lockdown were tested, however they did not have as much of an effect as expected.