A DATA MINING APPROACH TO PERFORMANCE REFINEMENT OF A HEURISTIC ALGORITHM
Elena Ikonomovska, Dejan Gjorgjevik
Abstract: This paper introduces a novel approach to performance refinement of a heuristic algorithm for combinatorial optimization. The proposed Adaptive Tabu Search (A–TS) algorithm introduces adaptive behavior in the traditional Tabu Search algorithm. The adaptive nature of this algorithm is based on two adaptive coefficients that drive the heuristic. Choosing appropriate values for these parameters has great impact on A–TS performance and accuracy. This article presents a novel approach towards improving the performance of A-TS utilizing data mining techniques for tuning the adaptive coefficients. The performance of A–TS was measured by applying it to the Quadratic Assignment Problem. Published results from other authors were used for comparison.