Shortest path deliveries using density-based clustering

Abstract

In e-commerce customers put orders online and goods are delivered to their destinations. The shortest path deliveries problem (SPD) is to compute a schedule for delivering the goods to all the destinations so that the total distance is minimal. In this paper we developed a new efficient delivery algorithm called DenCluSPD (Density-based Clustering for Shortest Path Deliveries). More specifically, we form a grid to hold the locations of the destinations and then create the clusters of various densities. By visiting the densest clusters first we speedup the performance and achieve the approximately optimal schedule. Our comprehensive experimentation results show that the new algorithm is indeed faster than the popular nearest neighbor first (NNF) algorithm and covers a shorter distance per destination initially.