|Abstract: ||對於公路運輸服務業而言，異常駕駛行為除直接導致潛在行車肇事或交通違規，提高業者經營風險，亦會降低行車服務品質，造成車輛無謂耗損徒增保養維修費用。如何針對駕駛員異常操作情形作出有效管理以及增進行車安全，是客運業者亟待解決的問題。隨著科技進步，智慧型運輸系統（Intelligent Transportation System, ITS）發展日益蓬勃，國內公路客運業者已漸由傳統「機械式行車紀錄器」改用數位式行車紀錄器。惟數位式行車紀錄器蒐集項目眾多與資料量相當龐大，如能夠有效分析與利用，對於公路客運業者行車安全、油耗與保修費用以及駕駛員管理將有所助益。|
For bus carriers, the abnormal driving behaviors will not only cause a higher risk of accident and traffic offence, but also deteriorate the vehicle worn-outs, which will cause bus service broken-down. Consequently, how to monitor and manage abnormal driving behaviors effectively and efficiently is an important issue to bus operators. With the progress of science and technology, Intelligent Transportation System is developed flourishingly. Today, many bus carriers have used digital tachographs to record the bus driving details. Particularly through data mining, the extraction of hidden predictive information from large databases will help us to find out and identify the relationship among abnormal driving behavior and driving safety, fuel consumption, and maintenance cost.
This study collected data from digital tachograph database, which include 61 buses, 210 drivers'' data. In the meanwhile, the vehicles related accident and traffic offence records, fuel consumption and maintenance cost data were also collected from June 1, 2004 to December 31, 2005. The first step of data mining is to confirm and define the variables and related threshold values of abnormal driving behaviors in order to extract the meaningful information. The data mining techniques were used in this study, such as multiple regression analysis and cluster analysis. Multiple regression models were developed to establish the empirical relationship among abnormal driving behaviors and driving safety, fuel consumption, and maintenance cost. The cluster analysis was applied to categorize the sample of drivers which have similar driving characteristics. The discriminate analysis was used to determine the driver''s level directly.
The findings of multiple regression models indicated that the emergent deceleration and speeding variables are the key determinants of the frequency of accident; speeding and abnormal operation of electromagnetic braking variables are the key determinants of the frequency of traffic offence; emergent deceleration and acceleration, long idle time of engine operation, abnormal operation of electromagnetic braking and abnormal engine rotation variables are the key determinants for fuel consumption and maintenance cost. The cluster analysis has classified the 210 drivers into 3 category levels: fair, bad and very bad. In this case, there are 199 drivers in fair level, 9 drivers in bad level, and 2 drivers in very bad level. According to the data mining results, this study proposed an integrated driver management solution. After evaluation, it is shown that applying driver management strategies such as re-education, rewards and punishments, on a monthly basis, the frequency of accident can be reduced by 13 times, frequency of traffic offence by 2 times and a cost of NT$745,480 for extra fuel consumption and maintenance cost are saved. The above savings could be used alternately for employees'' re-education and training, or to equip digital tachographs on the bus fleets.