Active Learning Based Rule Extraction for Regression

Advances in data mining have led to algorithms that produce accurate regression models for large and difficult to approximate data. Many of these use non-linear models to handle complex data-relationships in the input data. Their lack of transparency, however, is problematic since comprehensibility is a key requirement in many potential application domains. Rule-extraction algorithms have been proposed to solve this problem for classification by extracting comprehensible rule sets from the often better performing, complex models. We present a new pedagogical rule extraction algorithm for regression, based on active learning, which can be combined with any existing rule induction technique. Empirical results show that the proposed ALPA-R rule extraction method improves on classical rule induction techniques, both in accuracy and fidelity.