Abstract
Rule-based classifiers use a collection of high-quality rules to classify new data instances. They can be categorized according to the adopted classification strategy: Classifiers based on a single rule, and classifiers based on multiple rules. Many works were proposed in this field. However, most of them do not handle imperfect data. In this study, we focus on the issue of multi-rules-based classification for evidential data, i.e., data where imperfection is modeled via the belief functions theory. In this respect, we introduce a new algorithm called PWEviRC. This latter involves a two-level pruning technique to remove redundant and noisy rules. Finally, it applies the Dempster rule of combination to fuse the selected rules and make the final decision. To evaluate the proposed method, we carried out extensive experiments on several benchmark data sets. The performance study showed interesting results in comparison to existing methods.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have