Construction of opinion models for e-learning courses by rough set theory and text mining

Abstract

Extracting knowledge through the machine learning techniques in general lacks in its predictions the level of perfection with minimal error or accuracy. Recently, researchers have been enjoying the fruits of Rough Set Theory (RST) to uncover the hidden patterns with its simplicity and expressive power. In RST mainly the issue of attribute reduction is tackled through the notion of ‘reducts’ using lower and upper approximations of rough sets based on a given information table with conditional and decision attributes. Hence, while researchers go for dimension reduction they propose many methods among which RST approach shown to be simple and efficient for text mining tasks. The area of text mining has focused on patterns based on text files or corpus, initially preprocessed to identify and remove irrelevant and replicated words without inducing any information loss for the classifying models later generated and tested. In this current work, this hypothesis are taken as core and tested on feedbacks for e-learning courses using RST’s attribution reduction and generating distinct models of n-grams and finally the results are presented for selecting final efficient model.