Ai-related papers in the transportation field: Statistics and evolutionary trends

• AI-related publications in the transportation field are quantitatively analyzed. • The similarities and differences between various journals are discussed. • The competitions between China and US in this field are observed. • The statistical results of LDA and STM methods are compared. • Some unexpected patterns, including the decline of some keywords, are presented. It has become feasible to analyze research trends in the transportation field using NLP-based methods. However, existing studies mostly focused on the entire transportation discipline or on specific modes of transport, without specifically addressing the development of artificial intelligence (AI) technologies in transportation. To address this gap, this study collects AI-related publications in the transportation field and conducts a temporal evolution analysis. The selected journals include three categories: Transportation Research (TR) series, IEEE Transactions on Intelligent Transportation Systems (IEEE-ITS), and other representative Q1-Q2 transportation journals. Based on descriptive statistics, we apply LDA and STM method for topic modeling, and also use Vosviewer to analyze the network relationships among various elements. The results reveal several similarities among three categories: China ranks first in AI-related publication counts, followed by USA, with China surpassing USA in most cases between 2016 and 2019; Chinese institutions dominate the top positions in institutional statistics; the high-frequency terms in abstract texts are highly consistent across all journals; keyword clustering results consistently fall into three categories centered on machine learning, deep learning, and reinforcement learning. At the same time, notable differences are observed, including disparities in AI-related publication volumes, institutional affiliations, and topic modeling outputs. In addition, some unexpected patterns emerge: IEEE-ITS has seen a recent decline in publication volume, the keyword structure of IET-ITS has changed significantly, and some words once thought popular like “big data” and “data mining” have declined in prominence. Moreover, no new phenomena are identified when using STM, prompting further discussion.