Date of Award

2024-08-01

Degree Name

Master of Science

Department

Industrial Engineering

Advisor(s)

Sergio A. Luna Fong

Second Advisor

Ivonne Santiago

Abstract

Why despite all efforts to promote Electric Vehicles (EVs) as an alternative transportation method through strategies such as tax credits on unit purchasing or long-term environmental benefits communication, its market penetration has not reached the expected goals in the United States? Even though there have been important advancements in the EV technical perspective and financial EV purchasing incentives, the final EV customers still face barriers on their scenarios that do not allow them to purchase this type of contemporary transportation means. Not understanding their local barriers could be a mistake that could reduce EVA expectations in the country and keep, if not increase, the environmental impact, which could impact public health.The application of data collection methods such as surveys and the implementation of Machine Learning algorithms has been considered as the standard to run this type of analysis. However, the sample size bias (or imbalanced data) on the final dataset brings risks such as biased model performance, majority class overfitting, and misinterpretation of the results. This causes the stakeholders to deal with less prominent causes, wasting time and resources. To address this challenge, it is proposed the fusion of the Sentiment Analysis of social media posts, U.S. census attributes and U.S. charging stations datasets focused on three U.S. cities: Indianapolis, Indiana (IN); Salt Lake City, Utah (UT), and El Paso, Texas (TX) through Causal Inference. This approach results in defining which attributes impact the most public perception (or sentiment) towards EVA. The study results indicated that social media users' sentiment perceptions in the three cities were predominantly positive, followed by neutral. The diverse Electric Vehicle Adoption (EVA) conversation topics demonstrated empirically that the large volume of social media posts reflects the complexity of the topics discussed, such as EV Equity Awareness, EV Adoption Costs and EV Charging Infrastructure. In addition, to establish a causality between the sentiment perception and external variables, it was discovered that those have a scarce if not no correlation between them, having to rely on the observed social media posts information in the built of Bayesian Belief Network that allows to know the impact of these alternative variables in the sentiment polarity. The findings from this study give an addition to the decision-making process of stakeholders and policymakers regarding how to broadcast the EV transition message to communities with diverse backgrounds and develop strategies to achieve the EVA. Keywords â??Electric Vehicle Adoption, Social media data, Sentiment Analysis, Machine Learning, Feature Selection, Electric Vehicle, Charging station, Causal Inference, Charging Infrastructure.

Language

en

Provenance

Received from ProQuest

File Size

148 p.

File Format

application/pdf

Rights Holder

Jesus Alejandro Gutierrez Araiza

Download

Share

COinS