Predicting and Visualizing Consumer Sentiments in Online Social Media

With the rise of the Social Web, there is an explosive growth of user-contributed opinions toward products, services, and events on online social media platforms. These opinionated comments provide firms with unprecedented opportunities to tap into the collective consumer intelligence for enhancing marketing, product design, and other vital business functions. The main contribution of our research work is the design of a novel social media analytics framework on top of Apache Spark for predicting and visualizing consumers' opinion orientations based on their relationships with other consumers whose opinion orientations are known. In particular, we explore state-of-the-art collective classification (CC) algorithms for predicting consumers' opinion orientations. Based on real-world data collected from Sina Weibo, our experiments show that the proposed Gibbs sampling based CC algorithms which consider both a user's local features and their relational features outperform another approach that considers relational features alone.