Balanced K-Means Algorithm for Partitioning Areas in Large-Scale Vehicle Routing Problem

We present a new and effective algorithm, balanced k-means, for partitioning areas in large-scale vehicle routing problem (VRP). The algorithm divides two-stage procedures. The traditional k-means is used to partition the whole customers into several areas in the first stage and a border adjustment algorithm aims to adjust the unbalanced areas to be balanced in the second stage. The objective of partitioning areas is to design a group of geographically closed customers with balanced number of customers. The presented algorithm is specifically designed for large-scale problems based on decomposition strategy. The computational experiments were carried out on a real dataset with 1882 customers. The results demonstrate that the suggested method is highly competitive, providing the balanced areas in real application.