Metode Hybrid Grid Partition dan Rough Set untuk Pembangkitan Aturan Fuzzy pada Klasifikasi Data Set

Abstract

A fuzzy rule-based system with a grid-type fuzzy partition method to handle classification problems in low-dimensional patterns has shown the effectiveness of classification ability and very satisfactory interpretability, but this is not the case for high-dimensional data, the problem of increasing the number of rules still remains so that the classification system decreases. interpretability and classification accuracy. This research aims to develop a method for generating fuzzy rules for data set classification. The method developed is a hybrid method, namely the grid partition method and the rough set method, where the grid structure is formed using an adapted technique. The rough set method produces a set of reduct attributes based on variable precision or error rate. The data in the post-reduct information system table is reviewed in relation to the resulting redundancy pattern of condition attribute values and target attribute values, thereby reducing the number of attributes and the number of objects. Next, the fuzzy grid partition method generates fuzzy rules to obtain a collection of rules that can classify data sets. The research results show that the hybrid grid partition and rough set methods can generate the number of rules that do not increase exponentially and the classification accuracy level is higher, namely 83.33% compared to the fuzzy grid partition method with a classification accuracy level of 66.67%.