Case Study of Trend Mining in Transportation Research Record Articles

This study employs two topic models to perform trend mining on an abundance of textual data to determine trends in research topics from immense collections of unstructured documents over the years. This study collected data from the titles and abstracts of the papers published in Transportation Research Record: Journal of the Transportation Research Board, since 1974. The content of these papers was ideal for examining research trends in various fields of research because it contains large textual data. In previous studies, exploratory analysis tools such as text mining were used to provide descriptive information about the data. However, this method does not provide researchers with quantifications of the topics and their correlations. Furthermore, the contents examined in this study are largely unstructured, and therefore they require faster machine learning algorithms to decipher them. For these reasons, the research team chose to employ two topic modeling tools, latent Dirichlet allocation and structural topic model, to perform trend mining. This analysis succeeded in extracting 20 main topics, identified by keywords, from the data. The research team also developed two interactive topic model visualization tools that can be used to extract topics from journal titles and abstracts, respectively. The findings from this study provide researchers with a further understanding of research patterns within ever-evolving area of transportation engineering studies.