Classification Arabic Twitter User’s Insights Using Rough Set Theory

Abstract

Nowadays, people using social media from around the world to share their daily affairs. Arabic twitter for example is a platform where users read, reply, post which known ‘tweets’. Users trading their opinions on different trends that are not equal in important and differed based on their power and interest. Tweets can provide rich information to make decision. The main objective of this paper is to present a framework for making a valuable decision through analyzing social users' insights based on their proximity to a particular trend with highlights their power in this trend. Tweets are exceedingly unstructured that makes it difficult to analyze. Nevertheless, our proposed model differs from previous research in this field it gathered the use of supervised and unsupervised machine learning algorithms. The process of performing this work as follows: classifying users based on the degree of their closeness/interest utilizing Mendelow’s power/interest matrix, rough set theory to eliminate the features that may be found in user profiles to find minimal sets of data. The proposed model applied two attribute reduction algorithms on our dataset to determine the optimal number of reducts for improving decision making from the user replies. In addition to, unsupervised machine learning to group their replies into subcategories such as positive, negative, or neutral. The experimental evaluation shows that Johnson algorithm has reduced the user attributes by 71% than genetic algorithm that utilized in a classification model.