IdeaGraph Plus: A Topic-Based Algorithm for Perceiving Unnoticed Events

In the last few years, chance discovery as an extension of data mining has been proposed to capture rare but significant chances from a single document data for human decision making. Key Graph is a useful miner algorithm as well as a tool to discover chance candidates. On base of that, Idea Graph extended the concept of a chance to uncover more valuable chances. However, Key Graph and Idea Graph both fail to consider semantic relations among terms. In this paper, we propose an improved algorithm called Idea Graph plus which makes use of semantic information to enhance the performance of scenario construction using LDA topic model. Additionally, the term overlaps between sub-scenarios provide a thinking space for human to perceive unnoticed chances. An experiment demonstrates the superiority of Idea Graph plus by comparing with Idea Graph.