Travel-time Foresting Using Random Forest and Freeway ETC Data ─A Case of Taiwan Freeway between Taishan and Yangmei Toll Stations

• Main Authors: KUO PEI-HSIANG, 郭佩香
• Other Authors: 黃孝雲
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/56696212994431437522

碩士 === 輔仁大學 === 應用統計學研究所 === 97 === In the past, researchers commonly used machine learning systems such as artificial neural network (ANN) to predict travel time. Upon the embedded data driven concept, ANN utilizes existing materials to proceed with the real-time prediction. However, since vehicle's speed mobility is relatively high and the travel time must renew frequently, there still were problems against the method such as complicate variable selection and long operation time. This research aims to employ another new developed data driven method called Random Forest to forecast travel time. Furthermore, in order to obtain the capability of real-time prediction and improve the accuracy of forecasting, this research uses Freeway ETC materials between Taishan and Yangmei Toll Stations in 2007 as predict variables for the forecasting model. In the verification of practicability and feasibility of the proposed model, this research provides a performance comparison between the presented method, ANN and SVR to confirm the accuracy of the method. In that, RMSE is used as the forecasting error index.