Metaheuristics in Data Mining

INTRODUCTION The Metaheuristics are general strategies for designing heuristic procedures with high performance. The term metaheuristic, which appeared in 1986 for the first time (Glover, 1986), is compound by the terms: " meta " , that means over or behind, and " heuristic ". Heuristic is the qualifying used for methods of solving optimization problems that are obtained from the intuition, expertise or general knowledge (Michalewicz & Fogel, 2000). Nowadays a lot of known strategies can be classified as metaheuristics and there are a clear increasing number of research papers and applications that use this kind of methods. Several optimization methods that already existed when the term appeared have been later interpreted as metaheuristics (Glover & Kochenberger, 2003). Genetic Algorithms, Neural Networks, Local Searches, and Simulated Annealing are some of those classical metaheuristics. Several modern metaheuristics have succeeded in solving relevant optimization problems in industry, business and engineering. The most relevant among them are Tabu Search, Variable Neighbourhood Search and GRASP. New population based evolutionary metaheuristics such as Scatter Search and Estimation Distribution Algorithms are also quite important. Besides Neural Networks and Genetic Algorithms, other nature-inspired metaheuristics such as Ant Colony Optimization and Particle Swarm Optimization are also now well known metaheuristics..