Chance Discovery with Data Crystallization: Discovering Unobservable Events

摘要

The method of data crystallizing reveals the hidden structure by inserting dummy items corresponding to unobservable, i.e., hidden events, to the given incomplete and ill-structured data on past events. The existence of those hidden events and their location in the environment were visualized as a result of data crystallization. The method was evaluated by applying to (1) the simulated data using the 9/11 terrorist network and (2) test data provided by Dr. Bob Schrag via. It was further applied to two real-business domains: (1) redesigning surface inspection system (SIS), and (2) extracting the essence of flow of arguments in the negotiation logs for hypothetical two companies with respect to three business proposals.