Polaris: An Integrated Data Miner for Chance Discovery
Naoaki Okazaki, Yukio Ohsawa · 2003
KeyGraph, which is an algorithm as well as a tool for discovering rare or novel events, achieves successful outcomes in keyword extraction, earthquake prediction, genome analysis, sales promotion and marketing, questionnaire analysis, and so on. Through such case studies, Ohsawa proposed a double helical model of chance discovery in which humans and data-mining tools co-work; each progresses spirally toward creative reconstruction of ideas. However, existing data-mining tools do not connote architecture to promote chance discovery on the double helix model. We design Polaris, a new data mining system, that features graph representation of a source data as if a user observed a constellation of the data. Polaris accelerates a process of chance discovery by two strategies: it saves work and time for users to be close to their goal, i.e., what they want from the data; and it supports users to convince what they are actually thinking of, providing a way of analyzing their comments for an obtained graph.