CViz: An Interactive Visualization System for Rule Induction

We introduce an interactive visualization system CViz for rule induction The process of learning classi cation rules is visualized which consists of ve components preparing and visualizing the original data cleaning the original data discretizing numerical attributes learn ing classi cation rules and visualizing the discovered rules The CViz system is presented and each component is discussed Three approaches for discretizing numerical attributes including equal length equal depth and entropy based approaches are provided The algorithm ELEM for learning classi cation rules is introduced and the approaches to visualiz ing discretized data and classi cation rules are proposed CViz could be easily adapted to visualize the rule induction process of other rule based learning systems Our experimental results on the IRIS data Monks data and arti cial data show that the CViz system is useful and helpful for visualizing and understanding the learning process of classi cation rules