整合灰色遞迴類神經網路與模糊聚類之交通資訊與旅行時間預估

機構典藏 > College of Business and Management > Graduate Institute & Department of Transportation Management > Proceeding > Item 987654321/68950

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/68950

Title:	整合灰色遞迴類神經網路與模糊聚類之交通資訊與旅行時間預估
Other Titles:	Grey-based Recurrent Neural Network and Fuzzy Clustering for Traffic Information and Travel Time Estimation
Authors:	溫裕弘;Wen, Yuh-horng;李祖添;Lee, Tsu-Tian;卓訓榮;Cho, Hsun-Jung
Contributors:	淡江大學運輸管理學系
Keywords:	交通資訊;旅行時間預估;灰色遞迴類神經網路;模糊緊類;traffic infonnation;travel time estimation;grey-based recurrent neural networks for traffic forecasting;fuzzy clustering for data mining
Date:	2005-11
Issue Date:	2011-10-23 14:07:23 (UTC+8)
Publisher:	臺北市 : 中華民國運輸學會
Abstract:	本研究提出一整合預測與資料採礦之模式，對交通參數(流量、速率、佔有率或密度)資料與旅行時間進行預估，並透過資料採礦方法推測未來交通狀況。本研究提出一新的混合式類神經網絡模式，即灰色遞迴類神經網絡(grey-based recurrent neural network, G-RNN) ，整合灰色模式對資料隨機性與不確定性的有效處理，以及遞迴式類神經網路對時空特性的動態學習能力，進行交通參數預測與旅行時間預估。本研究進一步應用模糊c-means 眾類模式(Fuzzy-means clustering,FCM)基礎之資料採礦技術'針對大量的交通流量、速率與佔有率參數進行資料分析典型態辨識，以有效推測各類型交通狀況。本研究以高速公路實際偵測器資料進行範例分析，本研究G-RNN 模式之交通參數預測結果亦與灰色數列預測、其他顯神經網路模式作比較，比較結果以G-RNN 模式預測結果具較高準確度;G-RNN 旅行時間於20 秒-10 分鐘之預測結果準確度高(平均誤差約為3.17%) 。而模糊FCM 模式有效於大量交通資料中自動辨識、分群出三類車流狀況，並與過去相關車流研究結果相較驗證本研究模式之可行。本研究建構之整合架構能將交通資料進行即時而準且在之預測，並有效轉換為未來幾分鐘後之交通車流路況與預估路段旅行時間，即可適用於先進交通資訊系統之開發與應用。 This stuy presents a systematic process combining traffic forecasting and data mining models for trafffic information and travel time estimation. The hybrid grey-based recurrent neural network (G-RNN) was developed for traffìc parameter forecasting and travel time estimation. G-RNN integrates grey modeling into recurrent neural networks that is capable of dealing with both randomness and spatial-temporal properties in traffic data implicitly.Fuzzy c-means clustering model was developed for mining traffic flow-speed occupancy relationship then to extrapolate traffic information. Field data from Taiwan national freeway was used as a case study for testing the proposed models. Study results were shown that the G-RNN model is capable of predicting traffic parameters and travel times with α high degree of accuracy (90%-99%). The performance of G-RNN model is compared with grey time series model and recurrent networks. The application presents three clusters built from data and recognized three types of traffic conditions. Study results also showed feasibility of the method for advanced traffic information systems.
Relation:	中華民國運輸學會第二十屆學術論文研討會論文集 (第三冊)=Proceeding of the 20th annual conference for the Chinese institute of transportation v.3，頁885-905
Appears in Collections:	[Graduate Institute & Department of Transportation Management] Proceeding

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	302	View/Open

Loading...