Distance approximation to support customer selection in vehicle routing problems

Abstract Estimating the solution value of transportation problems can be useful to assign customers days for multi-period vehicle routing problems, or make customer selection decisions very fast (e.g., within an online environment). In this paper, we apply several regression methods predict total distance traveling salesman problem (TSP) and (VRP). We show that estimated fairly accurate using simple models only a limited number features. Besides features found in scientific literature, also introduce new classes spatial The model is validated on fictional case with different instances considering both backordering lost sales configuration, realistic involves dynamic waste collection city Amsterdam, Netherlands. For case, differences performance per instance type our save up $$25.3\%$$ 25.3 % compared heuristic approximation. cost function combines travel service level, reduce distances 17% well-known approximation while maintaining same level. Furthermore, benefits approximations combining offline learning frequent optimization.