Citation

Murugan M, Marisamynathan S. Int. J. Transp. Sci. Technol. 2023; 12(2): 428-446.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.ijtst.2022.03.008

PMID

unavailable

Abstract

As per statistics, two-wheeler (TW) alone shares the highest number of vehicle registrations in India, which develops the various transportation-related issues such as traffic conflicts, congestions, and pollutions. Electric two-wheelers (E-TW) are a better alternative to conventional two-wheelers because of their significant advantage in mitigating environmental impacts. But E-TWs are less attractive among road users due to unawareness of the benefits of E-TW. In addition, the traditional methods are less accurate in predicting users' mode choice behavior because of their limitations. Therefore, there is a need to conduct a study to understand the road user's willingness to adopt E-TW and find a suitable method for predicting mode choice behavior accurately. This study analyzes the Indian road users encouraging and discouraging factors to adopt E-TW and investigates the application of non-traditional models for estimating mode shift behaviour towards E-TW. Based on the literature review and expert opinion, a detailed questionnaire form was framed, and a total of 522 samples were collected from four states of India. The data findings show that Indian road users prefer TW compared to public transport, private four-wheeler, paratransit, and non-motorized transport because of its easy to ride, low maintenance, fast and convenient travel nature. The environmental concern of reducing air pollution and lower vehicle operating costs are significant factors that encourage E-TW adoption. However, the non-availability of charging infrastructure, lower speed, higher initial purchase cost, and lack of awareness about EVs are the significant discouraging factors in adopting E-TW in India. Further, Machine Learning (ML) methods were adopted to predict the mode shift behaviour from the fuel based TW to E-TW, and the results were compared with the Binary Logit (BL) method. The model results indicated that Support vector machine predicted the mode shift behavior with the highest accuracy rate compared with other methods such as Artificial Neural Network, K-Nearest Neighbor, Random Forest, and BL. The outcome of this study would help the transportation planner, EV manufacturers, researchers, and policymakers to understand the Indian user's preference to adopt E-TW.

Language: en

Keywords

Artificial neural network; Electric two-wheeler; Machine learning; Mode shift; Support vector machine