Two-step Classification Algorithm Based on Decision-Theoretic Rough Set Theory

Abstract

This paper introduces rough set theory and decision-theoretic rough set theory. Then based on the latter, a two-step classification algorithm is proposed. Compared with primitive DTRST algorithms, our method decreases the range of negative domain and employs a two-steps strategy in classification. New samples and unknown samples can be estimated whether it belongs to the negative domain when they are found. Then, fewer wrong samples will be classified in negative domain. Therefore, error rate and loss of classification is lowered. Compared with traditional information filtering methods, such as Naive Bayes algorithm and primitive DTRST algorithm, the proposed method can gain high accuracy and low loss.

DOI: http://dx.doi.org/10.11591/telkomnika.v11i7.2802