Making the Most of What You’ve Got: Using Models and Data to Improve Learning Rate and Prediction Accuracy

Prediction and classification in areas such as engineering, medicine, and applied expert systems often relies on two sources of knowledge: actual data and a model of the domain. Recent efforts in machine learning have developed techniques that take advantage of both sources, but the methods are often tied to particular types of models and induction techniques. We propose two general techniques that allow induction methods, C4.5 in our case, to take advantage of an available model.