Node-first causal network extraction for trend analysis based on web mining

Abstract

In this paper, we propose a node-first causal extraction method for trend analysis. Recently, it has become more important for business leaders, politicians and academics to understand broader and longer environmental trends because of the need to develop better strategies for dealing with current and future issues. Trend analysis is often utilized to identify key factors in political, economical, social and technological trends. We propose a web mining framework that can extract a causal network of key factors underlying macro trends related to a user's interest. The main idea can be described as "node-first" approach, which recursively identifies key factors relevant to a user's query, then verifies causal relations between key factors. As the result of experiment, we demonstrate high precision of key factor identification (P@100 = 0.76) and causality verifications (F-value = 0.74). © 2011 Springer-Verlag.