A new hybrid ant colony optimization algorithm for feature selection

Abstract

In this paper, we propose a new hybrid ant colony optimization (ACO) algorithm for feature selection (FS), called ACOFS, using a neural network. A key aspect of this algorithm is the selection of a subset of salient features of reduced size. ACOFS uses a hybrid search technique that combines the advantages of wrapper and filter approaches. In order to facilitate such a hybrid search, we designed new sets of rules for pheromone update and heuristic information measurement. On the other hand, the ants are guided in correct directions while constructing graph (subset) paths using a bounded scheme in each and every step in the algorithm. The above combinations ultimately not only provide an effective balance between exploration and exploitation of ants in the search, but also intensify the global search capability of ACO for a high-quality solution in FS. We evaluate the performance of ACOFS on eight benchmark classification datasets and one gene expression dataset, which have dimensions varying from 9 to 2000. Extensive experiments were conducted to ascertain how AOCFS works in FS tasks. We also compared the performance of ACOFS with the results obtained from seven existing well-known FS algorithms. The comparison details show that ACOFS has a remarkable ability to generate reduced-size subsets of salient features while yielding significant classification accuracy.