决策树课件课本课本课本(1)资料文档.pdfVIP

Some Basic Issues in CI/KDD/DM/ML/PR and Decision-tree algorithms College of Mathematics Computer Science Fuzhou University 2013-02-28 智能技术课程 A brief introduction to CI /KDD/DM/ML/PR • Acquire intelligence (Knowledge) from data by computation rather than by experts x1 x2 y • An extremely simple data set: 2-AND 1 1 1 1 0 0 Function? (analytic law): 0 1 0 y=x 1 x2 0 0 0 y=sgn (x1 + x2 - 1.5) Rules ? (descriptive law): If x1=0, then y=0; If x2 =0, then y=0; If x1=1 ∧ x2 =1, then y=1; Classifier ? (geometry / PR ): or more compact f(x , x ) = x + x - 1.5 1 2 1 2 If x1=1 ∧ x2 =1, then y=1, f(x , x ) = 0 as the separating line 1 2 otherwise y=0 2013-02-28 智能技术课程 Complete vs. Incomplete • 2-AND is complete in 2-dim Boolean space; • A data set in 3-dim Boolean space x1 x2 x3 y which is now incomplete: 1 1 1 1 1 0 0 0 Function? (analytic law): 0 1 0 0 y=x 1 x2 x3 0 0 0 0 y=sgn (x1 + x2 + x3 - 1.5) Rules ? (descriptive law): x1 x2 x3 y If x1=1 ∧ x2 ∧ x2 =1, then y=1, 1 1 1 1 otherw