Using Recurrent Neural Networks to Predict Station Level Demand in a Bike Sharing System

• Main Authors: Po-Chaun Chen, 陳柏全
• Other Authors: Jenq-Shiou Leu
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/12113214944629130433

碩士 === 國立臺灣科技大學 === 電子工程系 === 104 === Bike sharing systems have been widely applied to many cities and bringing convenience to their local citizens for short-term transportation. One of the biggest challenges in bike sharing systems is the bike rebalancing problem due to the unbalance of bikes distribution. In this thesis, we focus on station level prediction for each bike station. We propose four architectures based on recurrent neural networks and use only one model to predict both rental and return demand for every station at once which is very efficiency for online rebalance strategies used. Without taking the global level bike distribution into consideration, the MAPE/RMSLE of the sum over the demand of each station may be too high for rebalance strategies used even the MAE/RMSE are satisfied at station level. Our evaluation shows that the propose methods achieve not only satisfied results at station level, but also at global level on New York Citi Bike dataset.