Evaluating Electric Vehicle User Mobility Data using Neural Network based Language Models

The development of an efficient and reliable charging infrastructure is critical to reducing barriers to the adoption of electric vehicles (EVs). However, the availability of charging services can often be unreliable, even in dense urban centers. In this paper, the authors provide national evidence on how well publicly-owned and privately-owned charging stations are serving the needs of the rapidly expanding population of EV drivers in 651 core-based statistical areas in the United States. Natural language processing is applied to automatically classify 127,257 user reviews from the world's most popular EV charging station locator app. The authors classify user sentiments into positive and negative charging station experiences. To do this, they apply a convolutional neural network (CNN) and compare the model's classifications to those obtained from human raters. The authors find that the CNN-based classifer approaches human accuracy in this classification task (accuracy = 84.3%) and learns domain specific terms at the level of human experts. Contrary to expectations about the private provisioning of public charging services, the authors also find that privately-owned stations (e.g. hotels, commercial shopping areas or car dealerships), do not outperform publicly-owned stations (e.g. parks, municipal and government buildings, or transit centers) from the perspective of the consumer. Additionally, the authors find higher negative sentiment in dense urban centers, where issues of charge rage and congestion may be the most prominent.

• Supplemental Notes:
  • This paper was sponsored by TRB committee ADC80 Standing Committee on Alternative Transportation Fuels and Technologies.
• Corporate Authors:

  Transportation Research Board

  ,    
• Authors:
  • Alvarez, Kevin
  • Dror, Arielle
  • Wenzel, Emerson
  • Asensio, Omar Isaac
• Conference:
  • Transportation Research Board 98th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2019-1-13 to 2019-1-17
• Date: 2019

Language

• English

Media Info

• Media Type: Digital/other
• Features: Appendices; Figures; Maps; References; Tables;
• Pagination: 23p

Subject/Index Terms

• TRT Terms: Attitudes; Data analysis; Drivers; Electric vehicle charging; Electric vehicles; Mobility; Neural networks
• Uncontrolled Terms: Natural language processing (Computer science)
• Subject Areas: Data and Information Technology; Society; Vehicles and Equipment;

Filing Info

• Accession Number: 01698165
• Record Type: Publication
• Report/Paper Numbers: 19-05863
• Files: TRIS, TRB, ATRI
• Created Date: Mar 1 2019 3:51PM