MODIFICATIONS ON COMBINATION OF SPIRAL OPTIMIZATION METHOD WITH CLUSTERING TECHNIQUE AND PARALLEL IMPLEMENTATION OF THE PROGRAMS

• Main Author: Lestari, Susi
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/46713
• Institution: Institut Teknologi Bandung

The Spiral Optimization Algorithm (SOA) method is a metaheuristic search method inspired by spiral phenomena that occur in nature. The SOA method combined with clustering techniques can be used to find several solutions of multimodal optimization and the roots of a system of nonlinear equations in a single run of the program. The SOA method with clustering technique consists of 3 main phases those are the diversification phase (clustering), the intensification phase, and the final selection phase. In the intensification phase optimization is performed on each cluster independently. It is possible that clusters that do not contain solution but still be optimized. This research consists of 3 parts, modification 1 that is cluster modification in the clustering phase, modification 2 is the addition of threshold parameters before the intensification phase to reduce the number of clusters, and program implementation using parallel algorithms. Several benchmark equations have been tested in this study. The results show that algorithm with modification 1 gives better results. Algorithm with modification 2 which is limited only to the problem of finding the roots of the system of nonlinear equations quite well in reducing the number of clusters to be optimized but risky to do in certain systems of nonlinear equations cases. Programs run with parallel algorithms can increase computing speed by up to 10 times when compared to serial algorithms.