PrincipalComponentAnalysis(主成份分析).docVIP

Principal Component Analysis (主成份分析) 目的:Data reduction (summarizing data and reducing number of variables in regression analysis and clustering analysis) 步驟:將response variables(…,)轉換成(…,) kth Principal component linear combination of 看data的變化來自什麼樣的線性組合有最大變異，m k (1) =(…,) var-cov matrix , w.l.o.g assume = Find …) such that maximize var() over all linear combination of subject to ( direction of , not length of ) maximize subject to . Use Lagrange method L= satisfies 2 (-)=0 即 故 |-|=0 : largest eigenvalue of : eigenvector corresponding to first principal component Find m