DICE: A Discovery Environment Integrating Inductive Bias

Most of Knowledge Discovery in Database (KDD) systems are integrating efficient Machine Learning techniques. In fact issues in Machine Learning and KDD are very close allowing for a natural straightforward integration. However, there are specific problems related to KDD that require a specific approach to machine learning techniques integration. Overabundance of patterns and complexity of the discoveries is a central problem that we attempt to tackle. Our approach is to select several learning biases that are particularly relevant to KDD and to integrate them in a Discovery process. These learning bias are integrated into a KDD system, called DICE, that uses two Machine Learning Algorithm CHARADE and ENIGME. DICE offers an interface that allows the user to experiment with the learning bias. DICE is currently experimented on a medical database and a Chinese characters database.