Application of decision trees and multivariate regression trees in design and optimization
Babak Forouraghi · 1995
Induction of decision trees and regression trees is a powerful technique not only for performing ordinary classification and regression analysis but also for discovering the often complex knowledge which describes the input-output behavior of a learning system in qualitative forms;In the area of classification (discrimination analysis), a new technique called IDea is presented for performing incremental learning with decision trees. It is demonstrated that IDea's incremental learning can greatly reduce the spatial complexity of a given set of training examples. Furthermore, it is shown that this reduction in complexity can also be used as an effective tool for improving the learning efficiency of other types of inductive learners such as standard backpropagation neural networks;In the area of regression analysis, a new methodology for performing multiobjective optimization has been developed. Specifically, we demonstrate that muitiple-objective optimization through induction of multivariate regression trees is a powerful alternative to the conventional vector optimization techniques. Furthermore, in an attempt to investigate the effect of various types of splitting rules on the overall performance of the optimizing system, we present a tree partitioning algorithm which utilizes a number of techniques derived from diverse fields of statistics and fuzzy logic. These include: two multivariate statistical approaches based on dispersion matrices, an information-theoretic measure of covariance complexity which is typically used for obtaining multivariate linear models, two newly-formulated fuzzy splitting rules based on Pearson's parametric and Kendall's nonparametric measures of association, Bellman and Zadeh's fuzzy decision-maximizing approach within an inductive framework, and finally, the multidimensional extension of a widely-used fuzzy entropy measure. The advantages of this new approach to optimization are highlighted by presenting three examples which respectively deal with design of a three-bar truss, a beam, and an electric discharge machining (EDM) process.