Tactical Agility and Resilience Graduateship of Contemporary society of Business Practitioners (GSBP) programme Term: Clarence Ong (86596) Table of Articles I. II. III. 4. V. MIRE. VII.…...Read
Neurocomputing 55 (2003) 307 – 319 www.elsevier.com/locate/neucom
Financial period series forecasting using support vector machines Kyoung-jae Kim∗
Department of Information Systems, University of Organization Administration, Dongguk University, 3-26, Pil-dong, Chung-gu, Seoul 100715, South Korea Received twenty eight February 2002; accepted 13 March the year 2003
Abstract Support vector devices (SVMs) are promising methods for the prediction of ÿnancial timeseries because they use a risk function consisting of the empirical error and a regularized term which is produced from the strength risk minimization principle. This study is applicable SVM to predicting the stock cost index. Additionally , this analyze examines the feasibility of applying SVM in ÿnancial forecasting simply by comparing that with back-propagation neural systems and case-based reasoning. The experimental effects show that SVM provides a promising substitute for stock market conjecture. c 2003 Elsevier N. V. All rights appropriated. Keywords: Support vector equipment; Back-propagation nerve organs networks; Case-based reasoning; Financial time series
1 . Advantages Stock market prediction is regarded as a challenging activity of ÿnancial time-series prediction. There have been many studies using artiÿcial neural sites (ANNs) in this field. A large number of powerful applications have shown that ANN can be a very helpful tool to get time-series building and predicting . The early days of these research focused on application of ANNs to stock market prediction (for occasion [2, 6, 10, 13, 19, 23]). Recent exploration tends to hybridize several artiÿcial intelligence (AI) techniques (for instance [10, 22]). A lot of researchers tend to include novel factors in the learning process. Kohara et al.  incorporated previous knowledge to boost the ∗
Tel: +82-2-2260-3324; fax: +82-2-2260-8824. E-mail talk about: [email protected] kaist. ac. kr (K. -j. Kim).
0925-2312/03/$ - see front matter c the year 2003 Elsevier B. V. All rights reserved. doi: 10. 1016/S0925-2312(03)00372-2
K. -j. Kim / Neurocomputing 55 (2003) 307 – 319
performance of stock market conjecture. Tsaih ou al.  integrated the rule-based strategy and ANN to forecast the course of the S& P 500 stock index futures every day. Quah and Srinivasan  proposed an ANN share selection system to select stocks and options that are leading performers from the market and avoid picking under artists. They figured the collection of the recommended model perform better the portfolios of the standard model with regards to compounded real returns overtime, however,. Kim and Han  proposed a genetic methods approach to feature discretization and the determination of connection dumbbells for ANN to anticipate the stock price index. They advised that their approach lowered the dimensionality of the feature space and enhanced the prediction functionality. Some of these research, however , showed that ANN had some limitations in learning the habits because wall street game data has tremendous noise and complex dimensionality. ANN often demonstrates inconsistent and unpredictable performance on loud data. Yet , back-propagation (BP) neural network, the most popular nerve organs network unit, su ers from di culty in selecting a large number of controlling guidelines which include relevant input factors, hidden coating size, learning rate, impetus term. Lately, a support vector machine (SVM), a new neural network algorithm, was developed by Vapnik and his colleagues . Many classic neural network models got implemented the empirical risk minimization theory, SVM accessories the structural risk minimization principle. The previous seeks to minimize the mis-classiÿcation error or perhaps deviation by correct answer of the training data nevertheless the latter searches to minimize an upper certain of generalization error. In addition , the solution of SVM might be global the best while various other neural network models may possibly tend to get caught in a local ideal solution. Hence, overÿtting can be unlikely to happen with...
Referrals:  H. B. Achelis, Technical Analysis by A to Z, Probus Publishing, Chi town, 1995.  H. Ahmadi, Testability of the arbitrage charges theory by simply neural sites, in: Process of the Foreign Conference about Neural Networks, San Diego, FLORIDA, 1990, pp. 385 –393.  J. Chang, Sumado a. Jung, K. Yeon, J. Jun, D. Shin, L. Kim, Technical Indicators and Analysis Methods, Jinritamgu Creating, Seoul, mil novecentos e noventa e seis.  C. -C. Alter, C. -J. Lin, LIBSVM: a selection for support vector equipment, Technical Report, Department of Computer Scientific research and Info Engineering, Countrywide Taiwan University or college, 2001, Available at http://www.csie.edu.tw/∼cjlin/papers/libsvm.pdf.  J. Choi, Technical Indications, Jinritamgu Creating, Seoul, 1995.  L. H. Choi, M. T. Lee, Meters. W. Rhee, Trading S& P 500 stock index futures using a neural network, in: Process of the Annual Intercontinental Conference upon Artiÿcial Brains Applications about Wall Street, Ny, 1995, pp. 63–72.  D. L. Cooper, C. W. Emory, Business Analysis Methods, Irwin, Chicago, 1995.  They would. Drucker, Deb. Wu, V. N. Vapnik, Support vector machines intended for spam categorization, IEEE Trans. Neural Sites 10 (5) (1999) 1048–1054.  At the. Gi ord, Investor's Tips for Technical Analysis: Forecasting Price Action in the Market segments, Pitman Posting, London, 1995.  Sumado a. Hiemstra, Building structured nonlinear knowledge to predict currency markets returns, in: R. 3rd there�s r. Trippi (Ed. ), Mayhem & Nonlinear Dynamics inside the Financial Marketplaces: Theory, Proof and Applications, Irwin, Chicago, il, IL, 1995, pp. 163–175.  E. Kamijo, To. Tanigawa, Stock price pattern recognition: a recurrent nerve organs network approach, in: Procedures of the Worldwide Joint Seminar on Neural Networks, San Diego, CA, 1990, pp. 215 –221.
K. -j. Ellie / Neurocomputing 55 (2003) 307 – 319
 E. Kim, My spouse and i. Han, Genetic algorithms method of feature discretization in artiÿcial neural systems for the prediction of stock cost index, Experienced Syst. Appl. 19 (2) (2000) 125–132.  To. Kimoto, K. Asakawa, M. Yoda, M. Takeoka, Currency markets prediction system with flip neural network, in: Procedures of the Intercontinental Joint Meeting on Nerve organs Networks, Hillcrest, CA, 1990, pp. 1– 6.  K. Kohara, T. Ishikawa, Y. Fukuhara, Y. Nakamura, Stock cost prediction applying prior knowledge and neural networks, Int. J. Intell. Syst. Accounting Finance Control. 6 (1) (1997) 11–22.  S. Mukherjee, At the. Osuna, Farreneheit. Girosi, non-linear prediction of chaotic period series using support vector machines, in: Proceedings from the IEEE Workshop on Neural Networks intended for Signal Processing, Amelia Isle, FL, 1997, pp. 511–520.  M. J. Murphy, Technical Analysis of the Futures Marketplaces: A Comprehensive Guide to Trading Methods and Applications, Prentice-Hall, New York, 1986.  T. -S. Quah, B. Srinivasan, Enhancing returns upon stock purchase through neural network collection, Expert Syst. Appl. 17 (1999) 295–301.  F. E. H. Tay, M. Cao, Application of support vector machines in ÿnancial time series predicting, Omega 30 (2001) 309–317.  3rd there�s r. R. Trippi, D. DeSieno, Trading collateral index futures with a nerve organs network, J. Portfolio Take care of. 19 (1992) 27–33.  R. Tsaih, Y. Hsu, C. C. Lai, Forecasting S& G 500 share index futures with a crossbreed AI system, Decision Support Syst. twenty three (2) (1998) 161–174.  V. And. Vapnik, Statistical Learning Theory, Wiley, New york city, 1998.  I. They would. Witten, E. Frank, Info Mining: Useful Machine Study tools and Tactics with Java Implementations, Morgan Kaufmann Marketers, San Francisco, FLORIDA, 2000.  Y. Yoon, G. Swales, Predicting stock price efficiency: a nerve organs network strategy, in: Process of the 24th Annual Hawaii International Seminar on Program Sciences, Beautiful hawaii, 1991, pp. 156 –162.  G. Zhang, B. E. Patuwo, M. Sumado a. Hu, Predicting with artiÿcial neural networks: the state of the art, Int. J. Predicting 14 (1998) 35–62. Kyoung-jae Kim received his Meters. S. and Ph. M. degrees a manager Information Devices from the Graduate student School of Management with the Korea Advanced Institute of Science and Technology great B. A. degree from the Chung-Ang University or college. He is at present a faculty member of the Department of Information Devices at the Dongguk University. His research hobbies include info mining, know-how management, and intelligent real estate agents.