Augmented Visualization for Data-mining Models

This paper proposes a scheme called Augmented Model Visualization for Data Mining (AMV-DM), based on models of visual perception and interaction to represent and operate with data mining models. The scheme has at its core the use of complementary visualizations applied to data-mining (DM) model during the adjustment phase. These complementary views correspond to: a second descriptive technique of data mining, and an appropriate set of graphical artifacts. Defined metrics that measure the distance and similarity of components of a model and allow visual perception empirically data-analyst. AMV-DM is implemented through a prototype visual environment. As a case study explores a decision tree model and each of its nodes. Apply Self-Organizing Map technique on the decision tree (DT) model with a set of graphical artifacts. Two controlled experiments were carried out with 30 users. Preliminary results analysis allows obtaining empirical evidence of the usefulness of the proposed scheme.