大客車駕駛風險評估方法之構建與實證分析： 以A客運公司為例

Abstract

過去的交通安全研究著重於交通事故的環境因子、個人基本資料之風險統計分析，缺乏對於駕駛操作車輛機制與駕駛行為之了解與特性探討。即使有相關之探討，仍缺乏加入「駕駛工時」該項條件進行深入研究。因此本研究透過加入駕駛工時之評估條件，讓整體研究結果更加公正客觀，本研究使用SQL Server 進行ADAS車機資料的讀取與篩選，歸納出駕駛警示事件及各駕駛員在該月的駕駛工時，並設計專家問卷進行訪問，將得到的結果使用AHP層級分析法進行駕駛風險評估及警示事件之權重計算，得出各駕駛員的風險評分並進行排名，同時藉由Weka程式中的K-means分群法及其Python程式語言，將所有駕駛員分群，得出最安全到最不安全的駕駛員共有哪些，綜合上述結果，利用Power BI將資料視覺化，設計出駕駛員風險績效查詢頁面，方便客運公司的管理者可透過直覺化的查詢，進行駕駛員的篩選與該駕駛員是否安全的參考。

Past traffic safety research has focused on statistical analysis of environmental factors and basic personal data to assess the risks associated with traffic accidents. However, there has been a lack of understanding and exploration of the mechanisms and behaviors involved in vehicle operation and driving. Even when relevant discussions exist, there is still a lack of in-depth research on the condition of "driving hours." Therefore, this study aims to provide a more impartial and objective overall research result by incorporating the assessment of driving hours.

In this research, SQL Server is used to retrieve and filter ADAS (Advanced Driver Assistance Systems) vehicle data. Driver warning events and the driving hours of each driver in a given month are summarized. A questionnaire designed for experts is conducted to collect data. The results obtained are then subjected to Analytic Hierarchy Process (AHP) for driving risk assessment and calculation of the weights of warning events. This allows for the scoring and ranking of each driver's risk level. Additionally, the K-means clustering method in Weka and Python programming language is utilized to group all drivers and identify the safest and least safe drivers.

By integrating the above results, Power BI is employed to visualize the data and design a driver risk performance query page. This facilitates intuitive queries for managers of the passenger transport company, enabling them to filter drivers and make informed decisions regarding their safety.