Using Dependence Diagrams to Summarize Decision Rule Sets

Generating decision rule sets from observational data is an established branch of machine learning. Although such rules may be well-suited to machine execution, a human being may have problems interpreting them. Making inferences about the dependencies of a number of attributes on each other by looking at the rules is hard, hence the need to summarize and visualize a rule set. In this paper we propose using dependence diagrams as a means of illustrating the amount of influence each attribute has on others. Such information is useful in both causal and non-causal contexts. We provide examples of dependence diagrams using rules extracted from two datasets.