Metin siniflandirmada sezgisel arama algoritmalarinin performans analizi

Abstract

One of the most important problems in text categorization tasks is that the data space is very high dimensional which significantly diminishes the classification performance. Although, heuristic search algorithms are broadly used in many fields in the literature, they are not widely used in text categorization field. One important reason behind this fact is that these algorithms require high computational power and time to process the data for attribute selection purpose. In this study, a method to utilize such algorithms as a part of text categorization task is adopted and four popular heuristic search algorithms (Genetic Algorithm, Particle Swarm Optimization, Evolutionary Search and TABU Search) are tested. Obtained results show that heuristic search algorithms can be used effectively to increase the classification performance. Also, TABU algorithm has shown a slight performance advantage over the others.