MODL: A Bayes optimal discretization method for continuous attributes

Abstract

While real data often comes in mixed format, discrete and continuous, many supervised induction algorithms require discrete data. Efficient discretization of continuous attributes is an important problem that has effects on speed, accuracy and understandability of the induction models. In this paper, we propose a new discretization method MODL 1 , founded on a Bayesian approach. We introduce a space of discretization models and a prior distribution defined on this model space. This results in the definition of a Bayes optimal evaluation criterion of discretizations. We then propose a new super-linear optimization algorithm that manages to find near-optimal discretizations. Extensive comparative experiments both on real and synthetic data demonstrate the high inductive performances obtained by the new discretization method.

Predicting Dangerous Seismic Events in Coal Mines under Distribution Drift

“…In the MODL approach [4], the discretization is turned into a model selection problem and solved in a Bayesian way. Using a hierarchical prior distribution on the discretization parameters, the Bayes formula is applicable to derive an exact analytical criterion to evaluate the posterior probability of a discretization model.…”

Tagging Fireworkers Activities from Body Sensors under Distribution Drift

“…Figure 1 shows an example of discretization into three intervals, for the Sepal width input variable of the Iris dataset [8]. Total 57 57 36 Versicolor 34 15 1 Virginica 21 24 5 Setosa 2 18 In the MODL approach [9], the discretization is turned into a model selection problem and solved in a Bayesian way. First, a space of discretization models is defined.…”

Parsimonious Naive Bayes

“…Experiments demonstrate that even a simple equal width discretization brings superior performance compared to the assumption using a Gaussian distribution per class. In the MODL approach [7], the discretization is turned into a model selection problem and solved in a Bayesian way. First, a space of discretization models is defined.…”