ISSN (2005) vol. 3, no. 1, page 19~36 Inc,1999) Karnin f i E ( W ) E ( W ) f S w f i ( W W ) i W f W f f E( W ) W i i E( W ) W i W F f E ( )

Size: px

Start display at page:

Download "ISSN (2005) vol. 3, no. 1, page 19~36 Inc,1999) Karnin f i E ( W ) E ( W ) f S w f i ( W W ) i W f W f f E( W ) W i i E( W ) W i W F f E ( )"

戴性瞿
7 years ago
Views:

1 TEL: TEL: Graham Enron WorldCom Ball and Brown (Intrinsic Value) (Miller Freeman 19

2 ISSN (2005) vol. 3, no. 1, page 19~36 Inc,1999) Karnin f i E ( W ) E ( W ) f S w f i ( W W ) i W f W f f E( W ) W i i E( W ) W i W F f E ( ) ( ) ( u, w ) E w = w E w = 0 dw w I S S ^ ^ S f ^ N E Wij Sij = 1 ( n) Wij ( n) training testing W f i 0 ij wij Wij E Wij = η Wij f ^ N 1 2 W ij S ij = [ Wij ( n) ] generalization f i 0 η ( W ij W ij ) f N 1 2 W n 1 [ ηδ n ] ( ij j Y i f i 0 η W ij W ij ) ^ S Training Data 20

3 Class Classifier Testing Data ( P( v1 ),, P( vn ) I ( P( v = n 1 ),, P( vn )) P( vi ) log 2 P( vi ) i= 1 Decision Tree (Entropy) Boolean C i, i = 1,2,3 n function freq = ( C property i, S) S internal node freq ( C i, S ) S leaf node target class freq ( C i, S ) log ( ) 2 S implication sentences root path n freq( Ci, S) freq( Ci, S) propositional logic inf o( S) = log ( ) 2 i= 1 S S inf o( S) disorder S 1, S 2, S3, S m = n Si inf oa( S) inf o( Si) i= 1 S Information Theorem gain( A) = inf o( S) inf o ( S) A Quinlan gain 21

4 ISSN (2005) vol. 3, no. 1, page 19~36 γ ( x = n i ( k), x j ( k)) β kγ ( xi ( k), x j ( k)) k = 1 β k k n β k = 1 β k k= 1 Top to Bottom Grey Relational Step 1 Coefficient { P( X ); Γ} xi ( xi (1), xi (2), k, xi ( k)) X Step 2 i = 1 n, k = 1m γ ( xi ( k), x j ( k)) Step 3 x0 ( k) Step 4 x i (k) x0 ( k) x i (k) Step 5 min+ ς max γ ( x0 ( k), xi ( k)) = 0i ( k) + ς max x 0 Step 6 x0 3.2 Grey Relational Grade 1 n γ ( xi ( k), x j ( k)) = γ ( xi ( k), x j ( k)) n k = 1 22

5 DECISION TREE R * = ( ) 100 Beaver R* 23

6 ISSN (2005) vol. 3, no. 1, page 19~ Graham Warren Buffett 24

7 25

8 ISSN (2005) vol. 3, no. 1, page 19~36 26

9 27

10 ISSN (2005) vol. 3, no. 1, page 19~36 28

11 29

12 ISSN (2005) vol. 3, no. 1, page 19~36 30

13 31

14 ISSN (2005) vol. 3, no. 1, page 19~36 32

15 33

16 ISSN (2005) vol. 3, no. 1, page 19~36 6. Brown,P., The Impact of the Annual Net Profit Report on the Stock Market.,The Australian Accountant,1970,pp Edminster, R., An empirical test of financial ratio analysis for small business failure prediction, Journal of Financial and Quantitative Analysis,1972, No. 2, 8. Hamao, Y., Masulis, R., and Ng, V., Correlations in Price Changes and Volatility Across International Stock Markets, Review of Financial Studies, Vol.3,1990,PP Jang, G. S., Lai, F. P., Jiang, B. W., and Parng, T. M., Intelligent Stock Trading System with Price Trend Prediction and Reversal Recognition Using Dual Module Neural Networks, Journal of Applied Intelligence, Vol.3,1993, PP Levitt, M. E. and Vale, S.,Machine Learning for Foreign Exchange Trading, Neural Networks in the Capital Markets, 1995,PP Peter Temple, Magic Numbers for Stock Investors: How To Calculate the 25 Key Ratios for Investing Success, John Wiley & Sons, Vellido, A., Lisboa, P. J. G., and Vaughan, J., Neural Networks in Business a Survey of Applications ( 1992 ~ 1998 ), Journal of Expert System with Application, Vol.17,1999, PP.51 ~ Zhang, G., Patuwo, B. E., and Hu, M. Y., Forecasting with Artificial Neural Networks: The State of the Art, International Journal of Forecasting, Vol.14, 1998,PP Baba, N. and Kozaki, M., An Intelligent Forecasting System of Stock Price Using Neural Networks, IJCNN, Vol.1, 1992,PP Berry, M. J. A. and Linoff, G.,Data Mining Technique For Marketing, Sale, And Customer Support, Wiley Computer, Ball, R. and P.Brown, An Empirical Evaluation of Accounting Income Numbers,Joural of Accounting Research 6, 1968,pp Brown, The Impact of the Annual Net Profit Report on the Stock Market, The Australian Accountant,1970, pp Beaver, W. H., Financial Ratios as Predictors of Failures, Journal of Accounting Research, Vol.4,No. 1, 1966, pp

17 35

18 ISSN (2005) vol. 3, no. 1, page 19~36 Intelligent Multi layer selective Model and Application of Investment Strategy Ting-Cheng Chang Department of Commercial Technology and Management, Ling Tung University Kuan-Yu Lin Department of Finance, Ling Tung University Abstract The total numbers of the companies of Taiwan s security market, by the end of 2004, had already exceeded 1,200.How to select stocks and obtain much profit from the stock market is critical to investors. In this research, we attempt to use Data-mining technique to set up a multi layer selective model for providing recommendations on picking stocks. The multi layer model contains three steps in order to select critical financial ratio for training of Decision Tree, the first step is to evaluate the influence of the input variablesfinancial ratioson the output variablesfinancial structurefrom the component stocks of MSCI Taiwan index and financial crisis companies with sensitivity approach of Artificial Neural NetworksANN.The second step is to build up the standard procedure of selecting stocks with ROE and the principles from the training of Decision Tree. Finally, we find the order of selecting stocks with grey relation. Based on the empirical testing of all manufacturing industries of Taiwan from November 2000 to April 2005, the annual return of our portfolio was 89%.It defeated the return of the stock index. KeywordData MiningArtificial Neural NetworksSensitivity approachgrey Relational Analysis Decision Tree 36

Microsoft Word - A201202-493_1329751213.doc

Microsoft Word - A201202-493_1329751213.doc 5 10 15 20 25 BP 神经网络在中国创业板企业成长性预测研究 ** 孙静稳, 刘金平 ( 中国矿业大学管理学院, 江苏徐州 221116) 摘要 : 根据创业板企业的高科技和高成长性特点, 成为金融证券市场热门关注的对象, 其成长性研究是资本