An analysis of ridesharing trip time using advanced text mining techniques

Abstract

The time cost of ridesharing rental represents a crucial factor influencing users' decisions to rent a car. Researchers have explored this aspect through text analysis and questionnaires. However, the current research faces limitations in terms of data quantity and analysis methods, preventing the extraction of key information. Therefore, there is a need to further optimize the level of public opinion analysis. This study aimed to investigate user perspectives concerning travel time in ridesharing, both pre and post-pandemic, within the Twitter application. Our analysis focused on a dataset from users residing in the USA and India, with considerations for demographic variables such as age and gender. To accomplish our research objectives, we employed Latent Dirichlet Allocation for topic modeling and BERT for sentiment analysis. Our findings revealed significant influences of the pandemic and the user's country of origin on sentiment. Notably, there was a discernible increase in positive sentiment among users from both countries following the pandemic, particularly among older individuals. These findings bear relevance to the ridesharing industry, offering insights that can aid in establishing benchmarks for improving travel time. Such improvements are instrumental in enabling ridesharing companies to effectively compete with other public transportation alternatives.

Graphical abstract

Keywords

Ridesharing / Trip time / Topic modeling / Sentiment analysis / Twitter data