Analysis of Traffic Accidents Using Data Mining Techniques

• Main Authors: Yu-Ting Kuo, 郭玉婷
• Other Authors: Yung-Hsiang Hung
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/22297199409512964821

碩士 === 國立勤益科技大學 === 工業工程與管理系 === 102 === With the ever increasing demand for vehicles, traffic accidents gradually increase as the number of vehicles increases, resulting in the loss of lives and properties. The associated social cost is more difficult to estimate. The environment, time, and region influence the occurrence of traffic accidents, and life and property loss is expected to be reduced by improving traffic engineering, education, and administration of law and advocacy, thus, the waste of social cost may be reduced. This study used 2,471 traffic accident data of Taiping Dist. during January to December 2011, and employed Recursive Feature Elimination (RFE) of Feature Selection, Fuzzy Robust Principal Component Analysis (FRPCA), Back Propagation Neural Network (BPNN), and Logistic Regression (LR) to improve traffic accident forecast ability. Applies odds ratio of Logistic Regression explore the factors affecting the occurrence of accidents. The findings can help to prevent traffic accidents to secure lives and properties, provide reference for future road users, and for police to execute laws. The results showed that the performance of FRPCA-BPNN and FRPCA-LR combined with FRPCA in classification prediction is better than that of BPNN and LR. The separation facilities, road edge and lanes affect the driving behavior of drivers on the road. It often occurs the locations of road accident, such as forks, ramps, corners, street parked vehicles, significantly affect the driver's ability to judge.