Using Integer Programming to Rough Set Based Feature Selection: An Approach to Find All Reducts Respectively
Subject Areas : International Journal of Mathematical Modelling & Computations
1 - Department of computer science, Arak Branch, Islamic Azad university, Arak, Iran
Keywords: Rough set theory, Reduct, Core, Binary integer linear programming, Feature selection, Decision system.,
Abstract :
Rough set theory (RST) is an important tool to feature subset selection. One of the most critical and challenging issues in RST is finding reducts and Cores. Since most applied sciences involve high-dimensional descriptions of input features, much research has been done on dimensionality reduction. Feature Selection refers to the process of selecting the input features leading to the most predictable results. On the other hand, RST can be adopted to discover data dependencies and reduce the number of attributes in a data set using the data alone, requiring no extra information. Therefore, in this paper, we have proposed a simple approach for feature subset selection via binary integer linear programming (BILP). Optimal solutions to the result of this problem in reducts that lead to the features subset selection. All reducts are obtained in order from the smallest cardinality to largest cardinality. Also, to obtain optimal solutions for BILL, meta-heuristic algorithms such as genetic algorithm or standard method such as Branch and Bound method can be used. The steps of our approach are illustrated by an example.