Improved Fuzzy Rule Based Classification System Using Feature Selection and Bagging for Large Datasets

Classification process tries to classify any given test data to a set of predefined classes. The most widely used approach is fuzzy rule based classification system, where the learned model is represented as a set of IF-THEN rules. To deal with large datasets a Fuzzy Rule Based Classification System, chi-FRBCS algorithm is proposed. Based on the existing approaches a bagging method have been proposed. This uses the MapReduce framework for parallel processing of huge collection of data. But in the chi-FRBCS algorithm, the rules generated have high complexity and the accuracy of the classifier is not high. To reduce the complexity of the rules and maximize the accuracy of the classifier, bagging and feature selection methods are combined with chi-FRBCS algorithm. Bagging divides the datasets into n equal datasets and the feature selection choose the attributes which has high relevance with class attributes. By using chi-FRBCS algorithm in multi-class classification process the response time is low but the accuracy obtained is yet to be improved. Bagging and feature selection can be applied for multi class classification problem. The results demonstrate the accuracy of the system along with other state of art approaches.