Applying Data Mining Techniques for Traffic Accident Caused Casualties in Hualien

• Main Authors: Chen-ting Wu, 吳政庭
• Other Authors: Chih-Peng Chu
• Format: Others
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/63468551901874105820

碩士 === 國立東華大學 === 企業管理學系 === 104 === Traffic accidents amount in Taiwan, according to information provided by the Ministry of Interior found that since 2001, the country's total amount of traffic accidents increased year by year. Hualien is the largest city in Taiwan; however the regional public transport in Hualien is really inconvenient. How to reduce traffic accidents in Hualien become a matter. Above the description, this research chose data of Hualien traffic accident, arranged it and use traditional statistical analysis to find relationship between variables. Not only use traditional statistical analysis, but use C&RT, CHAID and Neural Network of data mining techniques to find out key variables and behaviors which are easily caused casualties in accidents. The accuracy of C&RT, CHAID models in this research are over 80%. Moreover, the accuracy of Neural Network is over 90%. But Neural Network didn’t produce rules to find behaviors easier to cause accidents, and it spent more time to build model. But C&RT and CHAID models can produce rules and CHAID spent least time and produced more rules. So this research chose CHAID model as best model. After analysis, this research found key variables were action status, vehicle type and license. From analysis, this research suggests government in Hualien should encourage people use public transportation and strengthen fundamental facility of public transportation, not only reduce traffic accidents, but also let tourists more convenient to use. Additionally, Hualien Traffic Police Corps should strengthen ban people illegal turning left, over limit, and have no proper license to reduce traffic accident.