Many search algorithms have been successfully employed in combinatorial optimization in logistics practice. This paper presents an attempt to weight the variable assignments through supervised learning in subproblems. Heuristic and exact search methods can therefore test promising solutions...