Simplifying Weighted Word Affinity Graph: an Approach Using Artificial Bee Colony Algorithm

An information retrieval system highly relies on document analysis/ retrieval system. It includes numerous processing stages such as feature extraction, semantic representation, dimensionality reduction and similarity measure. Semantic representation aids for providing a better description to the documents. However, the probability of getting increased dimension for semantic descriptors is high. Hence, dimensionality reduction method plays crucial role. Conventional dimensionality reduction methods such as Principle Component Analysis (PCA), Independent Component Analysis (ICA), etc entertains complex means of dimensionality reduction. In the literature, numerous classical optimization algorithms such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), etc. have been reported to solve the similar problem. However, valiant attempts have been made on deriving robust optimization over the traditional algorithms. Hence, we exploited Artificial Bee Colony Algorithm to solve the dimensionality reduction problem. In this paper, we first present a theoretical overview of mapping a dimensionality reduction problem to an optimization problem. Subsequently, we describe the procedural steps to solve the problem using artificial bee colony algorithm. This article is believed to be a context behind the experimental investigation on the performance of artificial bee colony algorithm, when attempting to reduce the dimension of weighted word affinity graph and to retrieve the information effectively.