Performance Comparison of SOM and ACO for Travelling Salesman Problem-Case Study on the Indonesia Palapa Ring Network

Abstract

Indonesia is an archipelagic country separated by the sea, so it has unique characteristics. Building a telecommunications network that optimally connects Indonesia's islands is essential. Currently, telecommunications networks in Indonesia are connected using the Palapa Ring Indonesia network. Several adjacent islands must specify the closest point to choose a neighbouring route. This paper uses a travelling salesman problem (TSP) that uses the Self-Organizing Map (SOM) and Ant-Colony Optimization (ACO) algorithms to determine the shortest route and optimal computation time. Paper modification has several parameters: number of iterations, neuron size, and space. The measurement results show that SOM is better at measuring the shortest route length and computational time than ACO. By setting the number of iterations below 500 and the M value as much as the number of neurons, the route obtained by SOM will be shorter with a shorter time.