A novel offline algorithm configuration method

Metaheuristic algorithms, which are developed to find near-optimal solutions for optimization problems within acceptable times, have specific parameters that have a significant effect on their performance. Although the selection of best-performing values for tunable algorithm parameters is a challenging and tedious task, it can lead to an effective and good-performing version of algorithms for optimization problems. In this thesis, we first give a formal explanation of the algorithm configuration problem and then survey the automated methods developed to handle this problem. Subsequently, we evaluated the performance of eight different sampling methods for the tuning of the parameters of the Differential Ant-Stigmergy Algorithm (DASA). Then, we proposed a novel algorithm configuration method based on the Latin Hypercube Hammersley Sampling (LHHS) and Fuzzy C-means Clustering methods (FCM), which were used for the first time for the algorithm configuration problem. Our experimental results show that the proposed tuning method outperformed existing state-of-the-art tuning methods in the two experiments and demonstrated competitive performance in the other two experiments. The most important result of our experiments is that not only the best parameter configuration but also other well-performed configurations found with the proposed method demonstrated competitive results with the best configuration found with other state-of-the-art algorithm configuration methods.