Data Mining in Stochastic Local Search

This chapter explores some stochastic local search heuristics that incorporate a data mining procedure. The basic idea of using data mining inside a heuristic is to obtain knowledge from previous iterations performed by a heuristic to guide the search in next iterations. Patterns extracted from good quality solutions can be used to guide the search, leading to a more effective exploration of the solution space. This survey shows that memoryless heuristics may benefit from the use of data mining by obtaining better solutions in smaller computational times. Also, some results are revisited to demonstrate that even memory-based heuristics can benefit from using data mining by reducing the computational time to achieve good quality solutions.