In this study, we develop guidance based on supervised machine learning to improve the metaheuristic algorithm for solving capacitated vehicle routing problem (CVRP). The CVRP is a combinatorial optimization problem that determines a least-cost set of routes from a depot node for a fleet of capacitated vehicles to meet the demands of a set of customer nodes. Mostly, algorithms for tackling this problem still solve from scratch, even for the same problem type. Meanwhile, leveraging historical data could prove invaluable for achieving efficient and effective solutions. Furthermore, the incorporation of machine learning (ML) offers the promise of real-time problem learning and guiding the algorithm towards efficient problem-solving. Therefore, this research aims to achieve two primary objectives: (1) to understand the connection between the quality of the solutions, their features, and the associated problem instances, and (2) to construct an efficient learning process consolidated with a robust optimization algorithm to solve the problems effectively.