Eco-driving for urban bus with big data analytics

Mu Chen Chen, Cheng Ta Yeh*, Yi Shiuan Wang

*Corresponding author for this work

Research output: Contribution to journalArticle

Abstract

Fuel consumption constitutes 20–30% of the operation cost of most bus companies. Consequently, reducing fuel consumption decreases operating costs and carbon emissions. Most previous studies adopted experimental methods to collect and analyze small data and focused on the influence of a single variable on fuel consumption. Therefore, the analytical results may not have appropriately reflected the operation requirements of the bus companies. Hence, this study obtains big data comprising of Telematics and operation records from an urban bus company and selects the relevant data according to several eco-driving aspects such as driving behavior, vehicle characteristics, driver characteristics, and weather. Subsequently, a decision tree, C5.0, is adopted to explore the relevant correspondence between variables that affect fuel consumption. Observing the analytical results, the variables of bus brand, bus age, bus weight, monthly passenger load, monthly salary, monthly working days, monthly overtime, and times of high-speed have relatively high influence on fuel consumption. Based on the results, therefore, several eco-driving recommendations of fuel consumption reduction are proposed. For the case of bus purchase, the urban bus company can cautiously consider bus brand, bus age, and bus weight. The company can also provide a friendly working environment with the reasonable monthly passenger load, monthly salary, working days in a month, and overtime to reduce the times of high-speed such that the fuel efficiency can be improved.

Original languageEnglish
JournalJournal of Ambient Intelligence and Humanized Computing
DOIs
StateAccepted/In press - 2020

Keywords

  • Big data
  • Decision tree
  • Eco-driving
  • Fuel consumption
  • Telematics
  • Urban bus

Fingerprint Dive into the research topics of 'Eco-driving for urban bus with big data analytics'. Together they form a unique fingerprint.

  • Cite this