Sentiment clustering using the unsupervised machine learning approach

These days, social media is crucial for disseminating opinions and remarks on various subjects. Users can utilize these remarks to help them process and make decisions in the future. Since the majority of the reviews are text or emojis, processing is necessary to acquire the necessary data. The reviews obtained from different source for processing is not labeled. When the reviews are labeled, the processing is done by using confusion matrix and the parameter are obtained from that for evaluation of the performance. But, when the reviews are not labeled, classification of the reviews are not possible as the original polarity is not available, which need to be compared with predicted polarity. Thus, unsupervised approach is adopted, where clustering approach is adopted. In this paper, three different unsupervised machine language are used namely, Affinity propagation, K-means, and Mini Batch K-means. The performance of the machine learning is checked by using performance evaluation parameters.