Data mining of IoT based sentiments to classify political opinions

Abstract

ABSTRACT In recent years, an exponential increase in the usage of social network services has been observed. These community services are typically used through different applications including personal computers, multiple applications of modern smartphones and wearable technologies. Proper identification and separation of different languages text, topic-based classification of text and classification of active users based on their published comments and posts are major challenges. In this research, our primary focus is to deal with English text collected through different IoT applications to analyse posts/comments to categorise people’s opinion in politics. We have developed an IoT framework model for collecting data from social media especially Facebook, preprocessed and clean data to be used for analysis, and separation of data based on different languages. Sentiment analysis techniques are used to detect polarisation of the individual user. The proposed system clustered IoT individuals based on their comments and posts and successfully detected political polarisation. The proposed approach obtained encouraging results with a precision of 66.7%, a recall of 71.4%, and an F-measure of 69.0% in the case of annotated data of 50 users and a precision of 75.0%, a recall of 87.1%, and F-measure of 80.6% in the case of annotated data of 100 users.