Simultaneous Feature Selection and Classification Using Fuzzy Rules

摘要

Classification of data plays a key role in the present conditions as most of the real world applications deal with large volumes of data. The problem of classification is defined as assigning a class to each data object in a way that is consistent with some observed data i.e., the Features, which we have about the problem. In reality these Features are of high dimensionality and designing a classifier for such data lead to more computational overhead thereby degrading the performance of the classifier and hence that leads to a problem called Curse of Dimensionality. In order to solve such problem different methods of data reduction have been used and managed to eliminate the redundancy and non-important features present in the data sets. Among them feature selection is a powerful approach of dealing with high dimensional data by selecting relevant features from data set. A rule-based system that is used to solve such problem should be designed such a way that the rules are generated from the Features that are extracted from the large volumes of data. The system can be designed by using a concept called Fuzzy logic. One of the main attractions of a fuzzy rule-based system is its interpretability which stops the increase in the dimensionality of the data. For high-dimensional data, the identification of fuzzy rules is also a big challenge. This work has described a flexible Feature Selection method based on Modulator Learning Algorithm and shows that the Modulator learning algorithm is capable of identifying better-quality feature subsets for most data sets than correlation feature selection (cfs) subset evaluator and a comparative study was carried using K-nearest neighbor, J48 and few existing classifiers of Weka. The effectiveness of the proposed method is demonstrated by carrying out experimental studies on benchmark datasets from the UCIML repository and one synthetic dataset.