Certain investigations on weight based Associative classification

Abstract

Association rule mining is an unsupervised mining method that can extract the newlinehighly relevant features from the given transactional dataset and present it as a rule with newlinean implication form. This representation can show the products that are purchased together. newlineOn the other hand, if a dataset consists of a class label, supervised learning such newlineas classification methods can be applied to build a classifier using a training dataset and newlinepredict the real-time data if the classifier provides acceptable test data accuracy. Combining newlinedifferent intelligent data analysis technique is a very active research area in data newlinemining. An associative classifier can be built by integrating classification and association newlinerule mining task. Recent research proves that an associative classifier can perform newlinebetter than the traditional classifiers. However, existing eager and lazy learning associative newlineclassifier methods do not consider feature weight in generating class association newlinerules. newlineThis research aims to improve associative classification performance by incorporating newlinethe weight in both eager and lazy learning associative classifier. Weight assignment newlineresults in a minimal number of high-quality weighted class association rules. The proposed newlinesystems have been experimented with benchmark datasets from the University of newlineCalifornia at Irvine Repository (UCI) on different evaluation metrics such as accuracy, newlineprecision, and recall. The proposed systems show an imporved performance compared newlinewith the existing systems. newline newline