Predict Topic Trend in Blogosphere

Graphical relationship among web pages has been used to rank their relative importance. In this paper, we introduce a way to leverage graphical relationship to capture characteristics of information diffusion. With this approach, we extract subgraphs by projecting the first few domains mentioning a topic onto the domain hyperlink graph and co-mention graph, and use machine learning method to construct a predictive model. Then we use a greedy feature selection algorithm to extract the most helpful features. Simulation result shows we are able to classify topic popularity with 78 % accuracy by observing even only first 5 or 10 mentioning domains