Ant colony optimization (ACO) is a well-explored meta-heuristic algorithm, among whose many applications feature selection (FS) is an important one. Most existing versions of ACO are either wrapper based or filter based. In this paper, we propose a wrapper-filter combination of ACO, where we introduce subset evaluation using a filter method instead of using a wrapper method to reduce computational complexity. A memory to keep the best ants and feature dimension-dependent pheromone update has also been used to perform FS in a multi-objective manner. Our proposed approach has been evaluated on various real-life datasets, taken from UCI Machine Learning repository and NIPS2003 FS challenge, using K-nearest neighbors and multi-layer perceptron classifiers. The experimental outcomes have been compared to some popular FS methods. The comparison of results clearly shows that our method outperforms most of the state-of-the-art algorithms used for FS. For measuring the robustness of the proposed model, it has been additionally evaluated on facial emotion recognition and microarray datasets.