Analysing Social Media Discourse on Electric Vehicles with Machine Learning

Yasin Özkara, Yasemin Bilişli, Fatih Serdar Yildirim, Fahrettin Kayan, Agah Başdeğirmen, Mehmet Kayakuş, Fatma Yiğit Açıkgöz

Social acceptance of electric vehicles is of great importance for environmental sustainability and economic development. This study aims to examine Turkish and English tweets about electric vehicles with sentiment analysis, text mining, and topic modelling techniques to reveal consumers’ electric vehicle purchasing behaviours, consumer perception and acceptance processes about electric vehicles, and social perceptions. The data was taken from the X platform, and high accuracy and F1 scores were obtained in both languages in the classification made with the deep learning-based LSTM model. The accuracy was 92.1% for English tweets and 96.7% for Turkish tweets. According to the sentiment analysis results, the perception of electric vehicles is generally positive in both languages. However, while the rate of neutral sentiment is higher in Turkish tweets, the rate of negative sentiment is higher in English tweets. This indicates that there is more criticism and debate about electric vehicles globally, while Turkish tweets have more neutral views on the subject. Word frequency analysis shows that positive comments about electric vehicles focus on economic and environmental advantages, while negative comments include concerns about charging time, battery life, and range concerns. The topic modelling identified three main themes related to electric vehicles: (1) reasons for being preferred by consumers and their purchasing tendencies, (2) the role of brands, (3) market developments and marketing strategies. In Turkish tweets, electric vehicle production, charging infrastructure, and consumer purchasing trends were at the forefront. In general, it is emphasised that charging infrastructure should be strengthened, battery performance should be improved, and costs should be reduced to accelerate the adoption of electric vehicles.