多获胜节点SOM及其在股票分析中的应用.pdfVIP

计算机研究与发展 ISSN N 111777TP J ournal of Comp ut er Research and Develop ment 45 (9) : 1493 1500 , 2008 多获胜节点 SOM 及其在股票分析中的应用 1 , 2 1 1 , 3 1 2 王丽敏 　梁艳春 　韩旭明 　时小虎 　李 　明 1 (吉林大学计算机科学与技术学院国家教育部符号计算与知识工程重点实验室 　长春 　1300 12) 2 (长春税务学院计算机科学与技术系 　长春 　130 117) 3 (长春工业大学信息传播工程学院　长春 　1300 12) (ycliang @j lu . edu . cn) MultiWinners SOMs and Their Appl ications to Stock Analysis Wan g Limin1 ,2 , Liang Yanchun1 , Han Xuming1 ,3 , Shi Xiaohu1 , and Li Ming2 1 ( Col leg e of Comp ute r S cience an d Technology , J i l i n Uni ve rs i ty , Key L aboratory of S y m bol Comp ut ation an d K now le d g e Eng i nee ri ng of t he M i nis t ry of E d ucation , Chang chun 1300 12) 2 ( D ep a rtment of Comp ute r S cience an d Technology , Chang ch un Tax ation Col leg e , Chang ch un 130 117) 3 ( I ns t i t ute of I nf orm ation S p rea d i ng Eng i neeri ng , Chang ch un Uni vers i ty of Technology , Chang chun 1300 12) Abstract 　In or der to enhance t he dynamic co mp etition and clu st erin g cap abilit y of selfor ganizing map ( SOM ) neural net wor k and imp rove t he p reci sio n of solution s , multiwinner s SOM mo del s are p rop o sed by ext endin g number s of winner s ba sed o n an un sup ervi sed SOM neural net wor k and by imp roving t he neighbor f unction an d weight f unction in t he net wor k . In addition , a t abumapping met ho d i s p rop o sed to avoid t hat t he same outp ut no de i s mapp ed by more t han one inp ut . The clu st erin g analy si s for t he stock i s u sed to examine t he eff ectivene ss of t he p ropo sed mo del s. The financial indexe s refl