Evolutionary response surfaces for classification: an interpretable model

摘要: Response surfaces are powerful tools for both classification and regression because they able to model many different phenomena construct complex boundaries between classes. With very simple expressions, response accurately solve difficult problems. Thus, the interpretability of results is interesting from point view expert, which provided by a useful information may be inferred. However, suffer major problem that limits their applicability. Even with low degree moderate number features, terms in extremely large. standard learning algorithms find problems efficiently obtain coefficients terms, risk overfitting high. To overcome this we present evolutionary two-class The use fitness function combines accuracy obtains accurate classifiers interpretable expert. obtained 20 UCI Machine Learning Repository comparable well-known more polynomial function.