This paper introduces the concept of chance discovery, i.e. discovery of an event significant for decision making. Then, this paper also presents a current research project on data crystallization, which is an extension of chance discovery. The need for data crystallization is that only the observable part of the real world can be stored in data. For such scattered, i.e. incomplete and ill-structured data, data crystallizing aims at presenting the hidden structure among events including unobservable ones. This is realized with a tool which inserts dummy items, corresponding to unobservable but significant events, to the given data on past events. The existence of these unobservable events and their relations with other events are visualized with KeyGraph, showing events by nodes and their relations by links, on the data with inserted dummy items. This visualization is iterated with gradually increasing the number of links in the graph. This process is similar to the crystallization of snow with gradual decrease in the air temperature. For tuning the granularity level of structure to be visualized, this tool is integrated with human's process of chance discovery. This basic method is expected to be applicable for various real world domains where chance-discovery methods have been applied.
Air-Temperature Response to Neighborhood-Scale Variations in Albedo and Canopy Cover in the Real World: Fine-Resolution Meteorological Modeling and Mobile Temperature Observations in the Los Angeles Climate Archipelago
