Classification of Uncontrolled Intersections Using Hierarchical Clustering

Abstract

This research helps define motorized level of service (MLOS) for different categories of uncontrolled intersections using mixed hierarchical clustering technique. Service and total delay has been considered for classifying MLOS and intersections, respectively. GPS is used to collect travel time and speed data for turning movements which are transformed to average delay values. Thirteen intersections from eight different cities in India form the dataset. Divisive followed by agglomerative clustering (DAC-HAC) algorithm is applied as a two-step process for obtaining the service and total delay ranges. Validation of clusters is performed based on Davies–Bouldin score, Calinski–Harabasz index and Silhouette gaps. Based on DAC-HAC, uncontrolled intersections are classified into six categories (Cat-I, II, III, IV, V and VI). Results indicate MLOS classes “D”, “E” and “F” have significantly higher service delay ranges as compared to Highway Capacity Manual “control” delay ranges indicating mixed traffic conditions. Most of the uncontrolled intersections under mixed traffic fall under Cat-IV, V and VI having higher total delay ranges (greater than 60 s/vehicle/approach). Finally, validation of the clustering results is done for geometric and roadside environmental features.