Highway characteristic Classification using TwoStep Clustering Algorithm: Methodology and Case Study in Korea

Abstract

This paper reports reasonable methodology of the optimal solution for the highway classification and discusses the results of experiments comparing four different clustering methods. A new concept of highway characteristic classification (HCC) and its methodologies were applied to identify traffic patterns in highway segments. The HCC consists of four different steps, such as data preprocessing, clustering, characterization, and classification. This study evaluated the performance of four clustering methods: Ward's minimum variance clustering method, K-means clustering method, Kohonen self-organizing map, and the Two-step clustering algorithm. The TwoStep clustering algorithm provides the best performance in term of within-group errors. The four clusters in the TwoStep clustering algorithm were determined the acceptable number of cluster. The highway schemes based on procedure of HCC are four designated area types as urban, suburban, rural, and recreational area.