Ant colony algorithm for text classification in multicore-multithread environment

Abstract

This paper presents about Ant Colony Algorithm (ACO) for Text Classification in Multicore-Multithread Environment in Artificial Intelligent domain. We had develop a software which assimilate concurrency concept to multiple artificial ants. Pheromone in ACO is the main concept used to solve the text classification problem. In regards to its role, pheromone value is changed depending on the solution finding that has been discovered at the pseudo random heuristic attempt in selecting path from text words. However, ACO can take up longer time to process larger training document. Based on the cooperative concept of ants living in colony, the ACO part is examined to work in multicore-multithread environment as to cater additional execution time benefit. In running multicore-multithread environment, the modification aims to make artificial ants actively communicate between multiple physical cores of processor. The execution time reduction is expected to show an improvement without compromising the original classification accuracy by the investment of trading on more processing power. The single and multicore-multithreaded version of ACO was compared statistically by conduction relevant test. It was found that the result shows a positive time reduction improvement.