非日常交通における車両感知器データ欠測を伴う路線の旅行時間予測ー機械学習によるアプローチー

• Item type: SIG Technical Reports(1)
• 公開日: 2020-09-01
• タイトル: 非日常交通における車両感知器データ欠測を伴う路線の旅行時間予測ー機械学習によるアプローチー
• タイトル / 言語: en
• タイトル: A Machine Learning Based Approach for Travel Time Prediction on Roadways with Missing Vehicle Detector Data in Unusual Traffic Flows
• 言語: jpn
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: 一般財団法人道路交通情報通信システムセンター
• 著者所属(英): en
• 著者所属(英): Vehicle Information & Communication System Center
• 著者名: 織田, 利彦
• 著者名(英): Toshihiko, Oda
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: VICS (Vehicle Information & Communication System) Center has provided drivers with useful traffic information. The travel time is usually generated based on the traffic flow data collected via vehicle detectors, and this service has offered excellent assistance and great convenience for drivers. Thus, the detectors are indispensable to observe and manage the traffic flows. However, owing to recent fiscal constraints, installation of detectors has been reduced significantly. Moreover, the renewal, repair and upgrade of the infrastructure has been delayed, which has led to degradation and malfunctioning of detectors. The degraded detectors have resulted in missing traffic flow data and lowered the quality of the service. Meanwhile, unusual traffic flows caused by large scale and special events have often occurred, and unexpected heavy congestion can be generated on the surrounding roads of events sites. To overcome these problems, we propose an innovative approach for compensation of missing detector data and prediction of travel time under unusual traffic flows by using a machine learning algorithm. Under the framework of the proposed approach, the training datasets are formed by sliding windows and generated by k-means clustering analysis, and a random forest model is applied to the prediction. In order to verify the effectiveness of the approach, we conducted an experimental study on arterial roads around Nakayama Racecourse on the days of the horse racing. The results show that our approach can predict travel time relatively accurately, which will lend strong support to enhance the information service provided by VICS Center.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: VICS (Vehicle Information & Communication System) Center has provided drivers with useful traffic information. The travel time is usually generated based on the traffic flow data collected via vehicle detectors, and this service has offered excellent assistance and great convenience for drivers. Thus, the detectors are indispensable to observe and manage the traffic flows. However, owing to recent fiscal constraints, installation of detectors has been reduced significantly. Moreover, the renewal, repair and upgrade of the infrastructure has been delayed, which has led to degradation and malfunctioning of detectors. The degraded detectors have resulted in missing traffic flow data and lowered the quality of the service. Meanwhile, unusual traffic flows caused by large scale and special events have often occurred, and unexpected heavy congestion can be generated on the surrounding roads of events sites. To overcome these problems, we propose an innovative approach for compensation of missing detector data and prediction of travel time under unusual traffic flows by using a machine learning algorithm. Under the framework of the proposed approach, the training datasets are formed by sliding windows and generated by k-means clustering analysis, and a random forest model is applied to the prediction. In order to verify the effectiveness of the approach, we conducted an experimental study on arterial roads around Nakayama Racecourse on the days of the horse racing. The results show that our approach can predict travel time relatively accurately, which will lend strong support to enhance the information service provided by VICS Center.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA11515904
• 書誌情報: 研究報告高度交通システムとスマートコミュニティ（ITS）; 巻 2020-ITS-82, 号 2, p. 1-6, 発行日 2020-09-01
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8965
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会