多项式与SVM预测模型理论分析及应用比较.pdfVIP

第31 卷 第9 期 合肥工业大学学报( 自然科 学版 ) V ol. 3 1 N o . 9 200 8 年9 月 JO U RN AL O F H E FEI U N IV E RSIT Y OF T E CH N OL O GY Sept. 2008 SV M 1 2 孙 林 , 杨世元 ( 1. , 230009; 2. , 230009) ( SVM ) ( LSSV M ) , ; SV M ; , SV M , , ; ; T M 7 15 A 100350 60 ( 200 8) 09148 105 Theoretic analysis and applicati on comparison between the polynomi al model and the support vector machine prediction model 1 2 SU N L in , Y A N G Shiy uan ( 1. Dept. of A p plied Ph y sics, Hef ei U niv ers ity of T echnology , H efei 230009, Chin a; 2. S chool of Inst rum ent S cien ce and Opt oel ect ronic E ngineering , H ef ei U niv ersit y of T echn ol ogy , H ef ei 230009, China) Abstract: T he standard suppor t vecto r machine( SV M) and it s adv anced for m- t he least squares sup port vect or machine ( L SSV M) are based on the pr inciple of st ru ct ural risk minimizat io n. T he predic t ion m odel based on t he L SSV M can su ccessfully solv e t he problems t hat t he polynom ial predict ion moded encounters . T he pap er analy zes t he sup erio rity of t he SV M model over t he po lynomial mo del theoret ically . Ex periment is also m ade. T he exp erim ent result show s th at t he SV M model is mor e ac cur at e and more robust in noise resist ance, and t hus m ore suit able f or pr edict ion. Key words: po lyno mial m odel; support vect or machine; annual pow er generat ion , SV M 2 , , ( L east SquaresSup ( Empir