Compression-Based Discretization of Continuous Attributes

Discretization of continuous attributes into ordered discrete attributes can be beneecial even for propositional induction algorithms that are capable of handling continuous attributes directly. Beneets include possibly large improvements in induction time, smaller sizes of induced trees or rule sets, and even improved predictive accuracy. We deene a global evaluation measure for discretizations based on the so-called Minimum Description Length (MDL) principle from information theory. Furthermore we describe the eecient algorithmic usage of this measure in the MDL-Disc algorithm. The new method solves some problems of alternative local measures used for discretization. Empirical results in a few natural domains and extensive experiments in an artiicial domain show that MDL-Disc scales up well to large learning problems involving noise.