Full-View: A Visual Data-Mining Environment

This paper presents a 3D user-centered interactive graphical environment dedicated to data mining. The aims of this environment are to involve the user in the data mining process (to use domain knowledge during the process), to improve comprehensibility (of both the data and the results of data mining algorithms), to improve interactivity and to use algorithms from various research areas: statistics, data analysis, visualization and machine learning. The environment is made of a set of bulletin boards where the graphical tools will be mapped; bulletin boards are predefined or can be user-defined. Several different visualization tools might be used in a single display, these tools are linked together to improve data comprehensibility. The tools available in the environment are both graphical and non-graphical tools, they can be used alone or in a cooperative way. One of these tools is more detailed: CIAD, a new graphical interactive decision tree construction algorithm that allows bivariate splits and so gives smaller trees (improving result comprehensibility). Its results are compared to existing decision tree algorithms. This environment can be used on any personal computer (it is based on open-source software and so, is platform independent) as well as on high performance graphical systems like reality centers.