Efficient Rule Set Generation using Rough Set Theory for Classification of High Dimensional Data

In this paper, a rough set theory (RST) based approach is proposed to mine concise rules from inconsistent data. The approach deals with inconsistent data. At first, it computes the lower and upper approximation for each concept, then adopts a learning from an algorithm to build concise classification rules for each concept satisfying the given classification accuracy. Lower and upper approximation estimation is designed for the implementation, which substantially reduce the computational complexity of the algorithm. UCI ML Repository datasets are used to test and validate the proposed approach. We have also used our approach on network intrusion dataset captured using our local network from network flow. The results show that our approach produceseffective and minimal rules and provide satisfactory accuracy over several real life datasets.