文档介绍：Generalization and Decision Tree Induction:
Efficient Classification in Data Mining
Micheline Kamber Lara Winstone Wan Gong Shan Cheng Jiawei Han
Database Systems Research Laboratory
School puting Science
Simon Fraser University, ., Canada V5A 1S6
g
f kamber, winstone, wgong, shanc, han ***@
Abstract 25, 26, 30]. A well-accepted method of classification is
the induction of decision trees [3, 25]. A decision tree
Efficiency and scalability are fundamental issues con- is a flow-chart-like structure consisting of internal nodes,
cerning data mining in large databases. Although classifi- leaf nodes, and branches. Each internal node represents
cation has been studied extensively, few of the known meth- a decision, or test, on a data attribute, and each outgoing
ods take serious consideration of efficient induction in large branch corresponds to a possible e of the test. Each
databases and the analysis of data at multiple abstraction leaf node represents a class. In order to classify an unlabeled
levels. This paper addresses the efficiency and scalability data sample, the classifier tests the attribute values of the
issues by proposing a data classification method which inte- sample against the decision tree. A path is traced from the
grates attribute-oriented induction, relevance analysis, and root to a leaf node which holds the class predication for
the i