Functional Domains Clustering of Autonomous Transportation Systems Based on Latent Dirichlet Allocation

As autonomous vehicles are coming to roadways, the transportation system will gradually become automated and connected, i.e., the current intelligent transportation systems (ITS) will evolve into autonomous transportation systems (ATS). As the basic element of ATS, function should be classified into functional domains, in order to better represent the architecture of ATS. The existing functional domain of ITS is mostly classified according to the experience of experts or practitioners, which lacks methodology and criterion. This paper explores how to categorize functional text into functional domain. In view of a large number of functional texts, the study uses Latent Dirichlet Allocation (LDA) topic model to extract text features, and inputs the weight of topic distribution into clustering algorithm to get 12 categories. Experiments show that the clustering results are basically in line with engineering requirements. This study provides a reference for text processing and function classification in the architecture of ITS.