Learning With Continuous Classes：（学习与连续的类）.pdfVIP

In Proceedings AI Adams Sterling Eds Singapore World Scientic LEARNING WITH CONTINUOUS CLASSES J R QUINLAN Basser Department of Computer Science University of Sydney Sydney Australia ABSTRACT Some empirical learning tasks are concerned with predicting values rather than the more familiar categories This pap er describ es a new system m that constructs treebased piecewise linear mo dels Four case studies are presented in which m is compared to other metho ds Intro duction One branch of machine learning empirical learning is concerned with building or revising mo dels in the light of large numb ers of exemplary cases taking into account typical problems such as missing data and noise Many such mo dels involve classication and for these learning algorithms that generate decision trees are ecient robust and relatively simple 17 Other tasks however require the learned mo del to predict a numeric value asso ciated with a case rather than the class to which the case b elongs For instance a case might b e the comp osition of an alloy and the asso ciated value the temp erature at which it melts Some researchers have attempted to use decision tree metho ds for value prediction by dividing the values range into small categories such as etc then using sys tems that build classication mo dels These attempts often fail partly b ecause algorithms for building decision trees cannot make use of the implicit ordering of such classes Several more eective learning metho ds are available for predicting real values The cart program 1 builds regression trees that dier f