Analysis of Freeway Accident Frequencies Using Data Mining Techniques

• Main Authors: Chen, Wen-Chieh, 陳文杰
• Other Authors: Chang, Li-Yen
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/74879880892044314555

碩士 === 國立嘉義大學 === 運輸與物流工程研究所 === 92 === The Poisson or negative binomial regression model has been employed to analyze vehicle accident frequencies for many years. However, these models have the pre-defined underlying relationship between dependent and independent variables. If this assumption is violated, the model could lead to erroneous estimation of accident likelihood. Data mining techniques which do not require any pre-defined underlying relationship between target (dependent) variable and predictors (independent variables) have been widely applied in the fields of business administration and engineering. Among the data mining techniques, the Classification and Regression Tree (CART) has been one of the most commonly employed techniques and shown to be a powerful tool in dealing with prediction and classification problems. This study collected the 2001-2002 accident data of National Freeway 1 in Taiwan. A CART model and a negative binomial regression model were employed to establish the empirical relationship between traffic accidents and highway geometric variables, traffic characteristics and environmental factors. The findings by CART indicated that the average daily traffic volume and precipitation days are the key determinants for freeway accident frequencies. By comparing the prediction performance between CART and negative binomial regression models, this study demonstrated that data mining techniques are alternative methods for analyzing freeway accident frequencies.