Financial time series forecasting with machine learning tech

Bond University ePublications@bond

Information Technology papers School of Information Technology 4-30-2010

Financial time series forecasting with machine learning techniques: A survey

Bjoern Krollner

Bond University , Bjoern_Krollner@e789ae323968011ca30091bc.au

Bruce Vanstone

Bond University , Bruce_Vanstone@e789ae323968011ca30091bc.au

Gavin Finnie

Bond University , Gavin_Finnie@e789ae323968011ca30091bc.au

This Conference Paper is brought to you by the School of Information Technology at ePublications@bond. It has been accepted for inclusion in Information Technology papers by an authorized administrator of ePublications@bond. For more information, please contact acass@e789ae323968011ca30091bc.au .Recommended Citation

Bjoern Krollner, Bruce Vanstone, and Gavin Finnie. (2010) "Financial time series forecasting with machine learning techniques: A survey" Paper presented at the European symposium on artificial neural networks: Computational and machine learning. Bruges, Belgium.Apr. 2010.

e789ae323968011ca30091bc.au/infotech_pubs/110

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

Financial Time Series Forecasting with

Machine Learning Techniques: A Survey

Bjoern Krollner, Bruce Vanstone, Gavin Finnie

School of Information Technology, Bond University

Gold Coast, Queensland, Australia

Abstract. Stock index forecasting is vital for making informed investment decisions. This paper surveys recent literature in the domain of machine learning techniques and artificial intelligence used to forecast stock market movements. The publications are categorised according to the machine learning technique used, the forecasting timeframe, the input variables used, and the evaluation techniques employed. It is found that there is a consensus between researchers stressing the importance of stock index forecasting. Artificial Neural Networks (ANNs) are identified to be the dominant machine learning technique in this area.

We conclude with possible future research directions.

1Introduction

Stock index prediction is an important challenge in financial time series prediction.

The stock market is subject to large price volatility which translates to high risks for holders of common shares. Portfolio persification permits the reduction of company specific risk, but the 2007/2008 financial crises highlighted the enormous effects of systematic market risk on portfolio returns. Derivative trading vehicles based on stock indices provide an effective means to hedge against systematic risk. In addition, they offer profit making opportunities for speculators. Determining more effective ways of stock index prediction is important for market participants in order to make more informed and accurate investment decisions.

This paper surveys recent literature in the domain of machine learning techniques and artificial intelligence used to forecast stock market movements. The main contribution of this survey is to provide researchers with a cohesive overview of recent developments in stock index forecasting and to identify possible opportunities for future research.

2Technologies Used

Machine learning techniques aim to automatically learn and recognise patterns in large amounts of data. There is a great variety of machine learning techniques within the literature which makes the classification difficult. This paper pides the literature into artificial neural network (ANN) based and evolutionary & optimisation based techniques.

Table 1 shows that variations of ANNs and hybrid systems are very popular in the recent literature. There is a clear trend to use established ANN models and enhance them with new training algorithms or combine ANNs with emerging technologies into hybrid systems.

25

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

Table 1: Reviewed papers classified by machine learning technique 3Forecasting Time-frame

Table 2 gives an overview of the different forecasting intervals used in the literature.

The prediction periods are categorised into one day, one week, and one month ahead predictions. Publications using multiple or different time-frame are listed under ’Multiple / Others’. Most papers make one day ahead predictions e.g. predic ting the next day’s closing price. However, being able to predict the stock index one day ahead does not necessarily mean that an investor can take advantage of this information in terms of trading profit, especially since the index itself cannot be traded. Surprisingly, only three publications [15, 22, 41] use data of actually tradable stock index futures for their studies.

Table 2: Reviewed papers classified by forecasting time-frame 4Input Variables

Selecting the right input variables is very important for machine learning techniques.

Even the best machine learning technique can only learn from an input if there is actually some kind of correlation between input and output variable.

Table 3 shows that over 75% of the reviewed papers rely in some form on lagged index data. The most commonly used parameters are daily opening, high, low and close prices. Also used often are technical indicators which are mathematical transformations of lagged index data. The most common technical indicators found in the surveyed literature are the simple moving average (SMA), exponential moving average (EMA), relative strength index (RSI), rate of change (ROC), moving average convergence / p ergence (MACD), William’s oscillator and average true range (ATR).

26

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

Table 3: Reviewed papers classified by input variables

5Evaluation Methods

In order to determine the effectiveness of a machine learning technique, a benchmark model is needed. A variety of evaluation methods is used in the literature. This survey categorises the evaluation models into the categories buy & hold, random walk, statistical techniques, other machines learning techniques, and no benchmark model.

Table 4 shows that the majority of authors use other machine learning techniques as a benchmark. This category consists of publications which perform a comparative analysis between two different models or use an established model and propose an improvement to that model. The proposed improved version is then compared to the original version.

Over 80% of the papers report that their model outperformed the benchmark model. However, most analysed studies do not consider real world constraints like

trading costs and slippage. 31 out of 46 studies use the forecast error as an evaluation

metric. This is a surprising finding since a smaller forecast error does not necessarily

translate into increased trading profits.

Table 4: Reviewed papers classified by evaluation models

27

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

6Conclusion

This paper has examined recent literature in the domain of machine learning techniques and artificial intelligence used to forecast stock market movements. The reviewed papers have been categorised according to the machine learning technique used, the forecasting time-frame, the input variables used, and the evaluation techniques employed.

In regards to the employed machine learning technique, there seems to be a trend to use existing artificial neural network models which are enhanced with new training algorithms or combined with emerging technologies into hybrid systems.

This finding indicates that neural network based technologies are accepted and suitable in the domain of stock index forecasting.

The surveyed forecasting time-frames revealed that the majority of publications tries to make one day ahead predictions using stock index data. It has been pointed out that for an investor it will be difficult to take advantage of this information, especially since the analysed literature does hardly examine any data of actually tradable derivatives.

Lagged index data and derived technical indicators have been identified as the most popular input parameters in the literature.

In summary, there seems to be a consensus between researchers stressing the importance of stock index forecasting and that the reported results are predominantly positive. Artificial Neural Networks (ANNs) have been identified as the dominant machine learning technique in this area.

The main finding of this survey is that there is a lack of literature examining if machine learning techniques can improve an investors’ risk-return tradeoff under real world constraints.

References

[1]Abraham, A., Nath, B. & Mahanti, P. K. (2001), Hybrid intelligent systems for stock market

analysis, in ‘Proceedings of the International Conference on Computational Science-Part II’,

Springer-Verlag, London, UK, pp. 337–345.

[2]Abraham, A., Philip, N. S. & Saratchandran, P. (2003), ‘Modeling chaotic behavior of stock indices

using intelligent paradigms’, Neural, Parallel Sci. Comput. 11(1 & 2), 143–160.

[3]Armano, G., Marchesi, M. & Murru, A. (2005), ‘A hybrid genetic-neural architecture for stock

indexes forecasting’, Information Sciences 170(1), 3–33.

[4]Bekiros, S. D. & Georgoutsos, D. A. (2008), ‘Direction-of-change forecasting using a volatility-

based recurrent neural network’, Journal of Forecasting 27(5), 407–417.

[5]Chen, A.-S., Leung, M. T. & Daouk, H. (2003), ‘Application of neural networks to an emerging

financial market: forecasting and trading the taiwan sto ck index’, Comput. Oper. Res. 30(6), 901–

923.

[6]Chen, Q.-A. & Li, C.-D. (2006), ‘Comparison of forecasting performance of ar, star and ann models

on the chinese stock market index’, Advances in Neural Networks 3973, 464–470.

[7]Chen, Y., Abraham, A., Yang, J. & Yang, B. (2005), Hybrid methods for stock index modeling, in

‘International Conference on Fuzzy Systems and Knowledge Discovery’, Springer Verlag, pp.

1067–1070.

28

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

[8]Chen, Y., Dong, X. & Zhao, Y. (2005), ‘Stock index modeling using eda based local linear wave let

neural network’, International Conference on Neural Networks and Brain 3, 1646–1650.

[9]Cheng, C.-H., Chen, T.-L. & Chiang, C.-H. (2006), ‘Trend-weighted fuzzy time-series model for

taiex forecasting’, Neural Information Processing 4234, 469–477.

[10]Chu, H.-H., Chen, T.-L., Cheng, C.-H. & Huang, C.-C. (2009), ‘Fuzzy dual-factor time-series for

stock index forecasting’, Expert Systems with Applications 36(1), 165–171.

[11]Chun, S.-H. & Kim, S. H. (2004), ‘Automated generation of new knowledge to support managerial

decision-making: case study in forecasting a stock market’, Expert Systems 21(4), 192–207.

[12]Collard, L. B. & Ades, M. J. (2008), Sensitivity of stock market indices to commodity prices, in

‘Proceedings of the 2008 Spring simulation multiconference’, The So ciety for Computer Simulation,

International, San Diego, CA, USA, pp. 301–306.

[13]de Faria, E., Albuquerque, M. P., Gonzalez, J., Cavalcante, J. & Albuquerque, M. P. (2009),

‘Predicting the brazilian stock market through neural networks and adaptive exponenti al smoothing

methods’, Expert Systems with Applications

[14]Fu, J., Lum, K. S., Nguyen, M. N. & Shi, J. (2007), ‘Stock prediction using fcmac-byy’, Advances in

Neural Networks 4492, 346–351.

[15]Hamid, S. A. & Iqbal, Z. (2004), ‘Using neural networks for forecasti ng volatility of s&p 500 index

futures prices’, Journal of Business Research 57(10), 1116–1125.

[16]Hanias, M., Curtis, P. & Thalassinos, J. (2007), ‘Prediction with neural networks: The Athens stock

exchange price indicator’, European Journal of Economics, Finance and Administrative Sciences 9,

21–27.

[17]Huang, S.-C. & Wu, T.-K. (2008), ‘Integrating ga based time-scale feature extractions with svms for

stock index forecasting’, Expert Systems with Applications 35(4), 2080–2088.

[18]Huang, W., Nakamori, Y. & Wang, S.-Y. (2005), ‘Forecasting stock market movement direction

with support vector machine’, Computers & Operations Research 32(10), 2513–2522.

[19]Huarng, K. & Yu, H.-K. (2005), ‘A type 2 fuzzy time series model for stock index forecasting’,

Physica A: Statistical Mechanics and its Applications 353, 445–462.

[20]Jaruszewicz, M. & Mandziuk, J. (2004), ‘One day prediction of nikkei index considering

information from other stock markets’, International Conference on Artificial Intelligence and Soft

Computing 3070, 1130–1135.

[21]Jia, G., Chen, Y. & Wu, P. (2008), ‘Menn method applications for stock market forecasting’,

Advances in Neural Networks 5263, 30–39.

[22]Kim, K.-J. (2004), ‘Artificial neural networks with feature transformation based on domain

knowledge for the prediction o f stock index futures’, Intelligent Systems in Accounting, Finance &

Management 12(3), 167–176.

[23]Kim, M.-J., Min, S.-H. & Han, I. (2006), ‘An evolutionary approach to the combination of multiple

classifiers to predict a stock price index’, Expert Systems with Applications 31(2), 241–247.

[24]Lee, T.-S. & Chen, N.-J. (2002), ‘Investigating the information content of non-cash-trading index

futures using neural networks’, Expert Systems with Applications 22(3), 225–234.

[25]Leigh, W., Hightower, R. & Modani, N. (2005), ‘Forecasting the new york stock exchange

composite index with past price and interest rate on condition of volume spike’, Expert Systems with

Applications 28(1), 1–8.

[26]Leung, M. T., Daouk, H. & Chen, A.-S. (2000), ‘Forecasting stock indices: a comparison o f

classification and level estimation models’, International Journal of Forecasting 16(2), 173–190.

[27]Liao, Z. & Wang, J. (2009), ‘Forecasting model of global stock index by stochastic time effective

neural network’, Expert Systems with Applications

29

ESANN 2010 proceedings, European Symposium on Artificial Neural Networks - Computational Intelligence and Machine Learning. Bruges (Belgium), 28-30 April 2010, d-side publi., ISBN 2-930307-10-2.

[28]Lu, C.-J., Lee, T.-S. & Chiu, C.-C. (2009), ‘Financial time series forecasting using independent

component analysis and support vector regression’, Decision Support Systems 47(2), 115–125.

[29]Majhi, R., Panda, G., Majhi, B. & Sahoo, G. (2009), ‘Efficient prediction o f stock market indices

using adaptive bacterial foraging optimization (abfo) and bfo based techniques’, Expert Systems with

Applications 36(6), 10097–10104.

[30]Majhi, R., Panda, G., Sahoo, G. & Panda, A. (2008), ‘On the development of improved adaptive

models for efficient prediction of stock indices using clonal-pso (cpso) and pso techniques’,

International Journal of Business Forecasting and Marketing Intelligence 1(1), 50–67.

[31]Ning, B., Wu, J., Peng, H. & Zhao, J. (2009), ‘Using chaotic neural network to forecast stock index’,

Advances in Neural Networks 5551, 870–876.

[32]Niu, F., Nie, S. & Wang, W. (2008), ‘The forecasts performance of gray theory, bp network, svm for

stock index’, International Symposium on Knowledge Acquisition and Modeling pp. 708–712.

[33]Pan, H., Tilakaratne, C. & Yearwood, J. (2005), ‘Predicting the australian stock market index using

neural networks exploiting dynamical swings and intermarket influences’, Journal of research and

practice in information technology 37(1), 43–55.

[34]Perez-Rodriguez, J. V., Torra, S. & Andrada-Felix, J. (2005), ‘Star and ann models: forecasting

performance on the spanish ibex-35 stock index’, Journal of Empirical Finance 12(3), 490–509.

[35]Roh, T. H. (2007), ‘Forecasting the volatility of stock price index’, Expert Systems with Applications

33(4), 916–922.

[36]Shen, J., Fan, H. & Chang, S. (2007), ‘Stock index prediction based on adaptive training and

pruning algorithm’, Advances in Neural Networks 4492, 457–464.

[37]Slim, C. (2004), ‘Forecasting the volatility of stock index re turns: A stochastic neural network

approach’, Computational Science and Its Applications 3045, 935–944.

[38]Stansell, S. R. & Eakins, S. G. (2004), ‘Forecasting the direction of change in sector stock indexes:

An application of neural networks.’, Journal of Asset Management 5(1), 37–48.

[39]Thawornwong, S. & Enke, D. (2004), ‘The adaptive selection of financial and economic variables

for use with artificial neural networks’, Neurocomputing 56, 205–232.

[40]Wang, W. & Nie, S. (2008), ‘The performance of several combining forecasts for stock index’,

International Seminar on Future Information Technology and Management Engineering 0, 450–

455.

[41]Witkowska, D. & Marcinkiewicz, E. (2005), ‘Construction and evaluation of trading systems:

Warsaw index futures’, International Advances in Economic Research 11(1), 83–92.

[42]Wu, Q., Chen, Y. & Liu, Z. (2008), Ensemble model of intelligent paradigms for stock market

forecasting, in ‘Proceedings of the First International Workshop on Knowledge Discovery and Data

Mining’, IEEE Computer Soci ety, Washington, DC, USA, pp. 205–208.

[43]Zapranis, A. (2006), ‘Testing the random walk hypothesis with neural networks’, Artificial Neural

Networks 4132, 664–671.

[44]Zeng, F. & Zhang, Y. (2006), ‘Stock index prediction based on the analytical center of version

space’, Advances in Neural Networks 3973, 458–463.

[45]Zhang, X., Chen, Y. & Yang, J. Y. (2007), Stock index forecasting using pso based selective neural

network ensemble, in ‘International Conference on Artificial Intelligence’, pp. 260–264.

[46]Zhu, X.,Wang, H., Xu, L. & Li, H. (2008), ‘Predicting stock index increments by neural networks:

The role of trading volume under different horizons’, Expert Syst. Appl. 34(4), 3043–3054.

30

本文来源：https://www.bwwdw.com/article/nwrl.html