Ekonometride Güncel Yöntemler ve Uygulamalar

EnglishTürkçe

BÖLÜM


DOI :10.26650/B/SS10.2021.013.19   IUP :10.26650/B/SS10.2021.013.19    Tam Metin (PDF)

GARCH, Yapay Sinir Ağları ve Destek Vektör Regresyonu Yöntemlerinin Karşılaştırılması: BİST Hizmetler Endeks Getirisi Örneği

Nimet Melis Esenyel İçen

Hisse senedi fiyatlarındaki oynaklığın tahmin edilmesi, hem bireysel hem kurumsal düzeyde yatırımcılar için önemini korumaktadır. Bu bakımdan uygun tahmin yönteminin belirlenmesi uzun zamandır araştırmacıların ve yatırımcıların üzerinde durduğu bir konu olmuştur. Çalışmada, klasik oynaklık modelleri ve yapay zeka modelleri kullanılarak Borsa İstanbul hizmetler endeks getirisinin öngörü performansının karşılaştırılması amaçlanmıştır. Klasik ve yapay zeka modellerine ek olarak hibrit modeller de oluşturulmuştur ve hibrit modellerin öngörü performansını arttırıp arttırmadığı incelenmiştir. Analize konu olan hizmetler sektörünün önemi, gelişmiş ve gelişmekte olan ülkelerde tarım ve sanayi sektörlerine kıyasla daha fazla gelişme göstermesidir. Bunun yanı sıra hizmet sektörü sektörel bazda yüksek bir paya sahip olduğundan ülkelerin gelişmişlik düzeyi ile alakalı olarak görülmektedir. Dolayısıyla hizmetler endeksinin modellenmesi ve tahmini önemli olmaktadır. Analizde kullanılan Borsa İstanbul hizmetler endeksi verileri TCMB Elektronik Veri Dağıtım Sisteminden alınmış olup, 05.01.2009- 13.11.2020 dönemini kapsayan günlük frekanstadır. Analiz sonucuna göre oluşturulan on model içerisinde öngörü performansı en yüksek modellerin hibrit modellerden DVR-EGARCH ve ELM-EGARCH olduğu görülmüştür.


Anahtar Kelimeler: Borsa İstanbul Hizmetler, GARCH modelleri, Yapay Sinir Ağları, Hızlı Öğrenen Makine, Destek Vektör Regresyonu
DOI :10.26650/B/SS10.2021.013.19   IUP :10.26650/B/SS10.2021.013.19    Tam Metin (PDF)

Comparison of Garch, Artificial Neural Networks, and Support Vector Regression Methods: the Case of the BIST Services Index Return

Nimet Melis Esenyel İçen

Measuring the volatility in stock prices is an important research concern at both individual and institutional levels. Individuals and businesses want to know the fluctuation of prices to avoid portfolio losses and to make profitable investments. Countries also try to understand the volatility in stock prices to determine investment policies at the international level. Consequently, establishing the appropriate estimation method has long been a topic of focus for researchers and investors alike.

Methods used in forecasting stock returns are generally classified as linear and nonlinear models. Linear models include autoregressive integrated moving average, exponential smoothing, and generalized autoregressive conditional heteroskedasticity (GARCH), among others. These models are mainly based on an assumption of linearity between variables that are normally distributed. Most of the studies that attempt to reveal the relationship between stock returns and financial and economic variables are based on simple linear regression assumptions. However, no evidence supports the assumption that this relationship is entirely linear. The significance of residual variance and the ability to explain this residual variance with nonlinear models makes it possible to obtain more reliable predictions using nonlinear models. Nonlinear models include those based on artificial intelligence, such as artificial neural networks (ANN), support vector machines, genetic algorithms, and particle swarm optimization. The assumption of linearity and normality may not be provided in the modeling of stock price movements. For this reason, artificial intelligence methods that have no restrictive assumptions have been recommended in the literature. These methods can also identify nonlinear structures, and have been shown to perform better empirically than statistical methods.

This study aims to compare the forecasting performance of the Borsa Istanbul (BIST) services index return applying classical volatility and artificial intelligence models. In addition to these models, hybrid models were also created to investigate whether hybrid models increase forecasting performance. The GARCH and exponential generalized autoregressive conditional heteroskedasticity (EGARCH), which are linear volatility models, as well as ANN and support vector regression (SVR), which are nonlinear models, will be used to predict stock returns. Hybrid models will also be created by combining ANN and SVR with GARCH and EGARCH models.

The services index is the stock market index calculated by considering changes in the stock prices of only services sector companies traded in the national market. There are 65 companies traded in the BIST services sector, and approximately 38% of these companies are in the “wholesale and retail trade, restaurants, and hotels” sector. The importance of the services sector is that it is more established in both developed and developing countries, as opposed to agricultural and industrial sectors. In addition, as the service sector has a high share on a sectoral basis, it is seen as related to a country’s development level, making the modeling and forecasting of the BIST services index an important consideration. The daily data used in the analysis is obtained from the Central Bank Electronic Data Delivery System covering the period 05.01.2009–13.11.2020. The results indicate that among the 10 models created and applied, those with the highest forecasting performance are the hybrid models SVR-EGARCH and extreme learning machine-EGARCH.


Anahtar Kelimeler: Borsa Istanbul Services, GARCH models, Artificial Neural Networks, Extreme Learning Machine, Support Vector Regression

Referanslar

  • Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics, 31(3), 307-327. https://doi.org/http://dx.doi.org/10.1016/0304-4076(86)90063-1 google scholar
  • Burges, C. J. C. (1998). A tutorial on support vector machines for pattern recognition. Data mining and knowledge discovery, 2(2), 121-167. google scholar
  • Chen, C. H. (1994). Neural networks for financial market prediction. Içinde Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN’94) (C. 2, ss. 1199-1202). google scholar
  • Durham, G. B. (2007). SV mixture models with application to S&P 500 index returns. Journal of Financial Economics, 85(3), 822-856. google scholar
  • Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation. Econometrica: Journal of the Econometric Society, 987-1007. google scholar
  • Enke, D., & Thawornwong, S. (2005). The use of data mining and neural networks for forecasting stock market returns. Expert Systems with applications, 29(4), 927-940. google scholar
  • Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The journal of Finance, 25(2), 383-417. google scholar
  • Hansen, J. V, & Nelson, R. D. (2002). Data mining of time series using stacked generalizers. Neurocomputing, 43(1-4), 173-184. google scholar
  • Hossain, A., & Nasser, M. (2011). Comparison of the finite mixture of ARMA-GARCH, back propagation neural networks and support-vector machines in forecasting financial returns. Journal of Applied Statistics, 38(3), 533-551. https://doi.org/10.1080/02664760903521435 google scholar
  • Huang, W., Nakamori, Y., & Wang, S.-Y. (2005). Forecasting stock market movement direction with support vector machine. Computers & operations research, 32(10), 2513-2522. google scholar
  • Jensen, M. C. (1978). Some anomalous evidence regarding market efficiency. Journal of financial economics, 6(2/3), 95-101. google scholar
  • Majhi, R., Panda, G., Sahoo, G., Panda, A., & Choubey, A. (2008). Prediction of S&P 500 and DJIA stock indices using particle swarm optimization technique. Içinde 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence) (ss. 1276-1282). google scholar
  • Mills, T. C. (1991). Nonlinear time series models in economics. Journal of Economic Surveys, 5(3), 215-242. google scholar
  • Nelson, D. B. (1991). Conditional heteroskedasticity in asset returns: A new approach. Econometrica: Journal of the Econometric Society, 347-370. google scholar
  • Oh, K. J., & Kim, K. (2002). Analyzing stock market tick data using piecewise nonlinear model. Expert Systems with Applications, 22(3), 249-255. google scholar
  • Ou, P., & Wang, H. (2010). Financial Volatility Forecasting by Least Square Support Vector Machine Based on GARCH, EGARCH and GJR Models: Evidence from ASEAN Stock Markets. International Journal of Economics and Finance, 2(1), 51-64. https://doi.org/10.5539/ijef.v2n1p51 google scholar
  • Package, T., Gosso, A. A., & Training, D. (2015). Package ‘ elmNN ’. google scholar
  • Perez-Cruz, F., Afonso-Rodriguez, J. A., Giner, J., & others. (2003). Estimating GARCH models using support vector machines*. Quantitative Finance, 3(3), 163-172. google scholar
  • Priestley, M. B. (1988). Non-linear and non-stationary time series analysis. nlns. google scholar
  • Tay, F. E., & Cao, L. (2001). Application of support vector machines in financial time series forecasting. Omega, 29(4), 309-317. https://doi.org/10.1016/S0305-0483(01)00026-3 google scholar
  • Ture, M., & Kurt, I. (2006). Comparison of four different time series methods to forecast hepatitis A virus infection. Expert Systems with Applications, 31(1), 41-46. google scholar
  • Wang, J.-Z., Wang, J.-J., Zhang, Z.-G., & Guo, S.-P. (2011). Forecasting stock indices with back propagation neural network. Expert Systems with Applications, 38(11), 14346-14355. google scholar
  • Wang, Y.-F. (2003). Mining stock price using fuzzy rough set system. Expert Systems with Applications, 24(1), 13-23. google scholar
  • Yao, J., Tan, C. L., & Poh, H.-L. (1999). Neural networks for technical analysis: a study on KLCI. International journal of theoretical and applied finance, 2(02), 221-241. google scholar
  • Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50, 159-175. google scholar
  • Zhong, X., & Enke, D. (2017). Forecasting daily stock market return using dimensionality reduction. Expert Systems with Applications, 67, 126-139. https://doi.org/10.1016/j.eswa.2016.09.027 google scholar


PAYLAŞ




İstanbul Üniversitesi Yayınları, uluslararası yayıncılık standartları ve etiğine uygun olarak, yüksek kalitede bilimsel dergi ve kitapların yayınlanmasıyla giderek artan bilimsel bilginin yayılmasına katkıda bulunmayı amaçlamaktadır. İstanbul Üniversitesi Yayınları açık erişimli, ticari olmayan, bilimsel yayıncılığı takip etmektedir.