Prediction of brand stories spreading on social networks

Abstract

Online social network is a major media for many types of information communication. Although the primary purpose of social networks is to connect people, they are more and more used in online marketing to connect businesses with customers as well as to connect customers amongst themselves. Brand stories generated by consumers or businesses can be easily and widely spread. As a result, those stories have a huge influence on the marketplace and indirectly affect the brand success. Understanding and modeling how a piece of information is spread on social media and its spreading level are crucial for business managers; not only to understand the information diffusion, but also for them to better control it. In this paper, we aim at developing models in order to predict the spread of brand stories on social networks, both in term of spreadability and spreading level. We applied several machine learning algorithms using three categories of features based on user-profile, temporal, and content of tweets. Experimental results on three tweet collections about brand stories reveal that our model significantly improves the prediction accuracy by about 4% compared to the related work.