Integrated method of compromise-based ant colony algorithm and rough set theory and its application in toxicity mechanism classification

Attribute discretization and reduction are two key issues in rough set theory. However, almost all previous studies consider them as two separate steps, which can not capture an inherent relationship between them. In this paper, a bi-objective optimization problem is constructed for simultaneous attribute discretization and reduction. A novel compromise-based ant colony algorithm (CACA) for simultaneously solving attribute discretization and reduction is proposed, which adopts a distance metric to stepwise approach the ideal solution. To improve efficiency of the proposed method, both the cut information and attribute information are adopted to dynamically calculate heuristic information, and a local search strategy is also embedded. The grade of nature spearmint essence (NSE), wine and glass classification problems are used as three test datasets to demonstrate the validity of the proposed CACA. Furthermore, the proposed method is applied to two toxicity mechanism classification problems: the classification of three narcosis mechanisms of aquatic toxicity for 194 organic compounds and the classification of four action modes of 221 phenols. The obtained results illustrate that the proposed method has better prediction performance than linear discriminant analysis, radial basis function neural network and support vector machine.