Sentiment on social interactions using linear and non-linear clustering

Social media analytics play a major role in e-commerce for extracting the useful information of a product or service. Opinion mining has become the key process of social media analytics. Twitter is a big online social activity platform where millions of people share their opinions. In this paper two clustering techniques, k-means and DBSCAN, are applied to an annotated Twitter dataset in order to evaluate use of clustering for detecting different types of sentiment. Results are very encouraging for DBSCAN but less useful for k-means.