The Empirical Evaluation of Models Predicting Bike Sharing Demand

Abstract

Most bike sharing system has an imbalance problem in certain time zones and certain rental stations where bicycles are insufficient or overloaded. So, a demand forecasting model is required to solve this problem. In this paper, we evaluate the performance applying the machine learning, neural network model with the bicycle demand dataset collected from the bicycle sharing system in actual operation in order to develop a model that predicts bicycle demand information for choosing a proper demand forecasting model. From the results, the neural network models outperform the machine learning models.