An incremental rough set approach for faster attribute reduction

Abstract

In the view of performance improvement in machine learning algorithms it is essential to feed them with minimal and most relevant features. Feature selection is one of the evident preprocessing step followed by most of the learning algorithms for choosing the relevant features towards reducing the dimensionality of the dataset as well as to improve the classification accuracy. Among various feature selection techniques, Rough Set Theory has its own major contributions in feature selection domain. However, the conventional Rough Set based Feature Selection (RSFS) procedure takes up to O(kn2) time and space complexity, where k is the number of objects and n is the total number of attributes, which is further reduced to O(kn). In this paper, an incremental approach of RSFS is proposed to reduce the time complexity of the algorithm to O(k log n) time. Here the indiscernibility relation is estimated in an incremental fashion rather than calculating them in conventional way, hence the overall attribute reduction time is reduced significantly when compared to conventional roughest approach. The performance of the proposed Incremental Rough Set (InRS) approach is evaluated with QuickReduct feature selection algorithm. The simulation results indicate that the InRS approach is able to find the reduct with minimal time complexity than the existing rough set methods.