BİST 100 Endeksindeki Volatilitenin Zaman Serileri İle Analizi: Fbprophet ve LSTM Modeli Karşılaştırması

Geçmişteki verilere dayanarak gelecek hakkında tahminler yapmak analitik finanstaki en önemli konulardan birisidir. Son dönemde gelişen derin öğrenme yaklaşımları ve makine öğrenmesi modelleri bu alana olan ilgiyi arttırmıştır. Bu yaklaşımlardan birisi olan zaman serileri ile belirli frekanstaki değişimler tahmin edilmeye çalışılmaktadır. Bu çalışmada BIST-100 endeksine ait verileri tahmin edebilmek için LSTM (Long Short-Term Memory) ve Fbprophet (Facebook Prophet) yöntemleri kullanılmıştır. Düzensiz davranışlara sahip borsa endekslerinin tahmin edilmesi karmaşık bir iştir ancak geliştirilen yeni algoritmalar ile fiyat tahminleri daha öngörülebilir hale gelebilmektedir. Araştırma yüksek volatiliteye sahip 2021-01-01 ile 2021-12-31 arasındaki endeks verileri üzerinden gerçekleştirilmiştir. Kullandığımız modellerin değerlendirme kriterleri MAE (ortalama mutlak hata), MSE (ortalama kare hatası) ve RMSE (kök ortalama kare hatası)’dır. Çalışma sonucunda düşük hata oranları ile LSTM modelinin Fbprophet modelinden daha başarılı olduğu tespit edilmiştir.

Anahtar Kelimeler:

BİST 100 Endeksi, Makine Öğrenmesi, Derin Öğrenme, LSTM, Fbprophet.

Analysis of Price Volatility in BIST 100 Index With Time Series: Comparison of Fbprophet and LSTM Model

Making predictions about the future based on past datasets is one of the most important issues in analytical finance. Recently developed deep learning approaches and machine learning models have increased the interest in this field. One of these approaches, time series, is trying to predict the changes in a certain frequency. In this study, LSTM (Long Short-Term Memory) and Fbprophet (Facebook Prophet) methods were used to estimate the data of BIST-100 index. Predicting stock market indices with erratic behavior is a complex task, but with the new algorithms developed, price predictions can become more predictable. The research was carried out on the index data between 2021-01-01 and 2021-12-31, which has high volatility. The evaluation criteria of the models we used are MAE (mean absolute error), MSE (mean square error) and RMSE (root mean square error). As a result of the study, it was determined that the LSTM model was more successful than the Fbprophet model with a low error rates.

Keywords:

BIST 100 Index, Machine Learning, Deep Learning, LSTM, Fbprophet.,

PDF

___

Ma, R., Zheng, X., Wang, P., Liu, H. and Zhang, C. (2021). The prediction and analysis of COVID-19 epidemic trend by combining LSTM and Markov Method. Scientific report, (11), Number 17421.
Nguyen, H. D., Tran, K. P., Thomassey, S and Hamad, M. (2021). Forecaing and anormaly detection approaches using LSTM and LSTM Autoencoder techniques with the application in supplu chain management. İnternational Jornal of Information Management. V. 57, n, 102282.
Lu, H., Ge, Z., Song, Y., Jiang, D., Zhou, T. And Qin, J. (2021). A temporal-aware LSTM enhanced by loss-switch mechanism for traffic flow forecasting. Neurocomputing. Volüme 427, p:169-178.
Kwon, D., Kim, J., Heo, J., Kim, C. And Han, Y. (2019). Time Series Classification of Cryptocurrency Price Trend Based on a Recurrent LSTM Neural Network. Journal of Information Processing System. 15(3), pp. 694-706. DOI: 10.3745.JIPS.03.0120.
Aditya P., B., Dvareddy, S., Hegde, S. and Ramya, B., S. (2021). A Time Series Cryptocurrency Price Prediction Using LSTM. Emerging Research in Computing Information Communication and Applications. V. 790, pp 653-662.
Andi, H., K. (2021). An accurate Bitcoin Price Prediction Using Logistic Regression with LSTM Machine Learning Model. Journal of Soft Computing Paradigm, 3(3), 205-217. Doi:10.36548/jscp.2021.3.006
Baek, Y. and Kim, Y. (2018). ModAugNet: A new forecasting framework for stock market index value with an overfitting prevention LSTM modüle and a prediction LSTM modüle. Expert System with Application. 113(15). Pp 457-480.
Rana, M., R., Rahman, F., Faysal, J. and Rahman A. (2021). An Effective Prediction on COVID-19 Prevalence for India and Japan using Fbprophet Model. Asian Journal of Research in Computer Science. 11(2): 16-28. ISSN: 2581-8260
Chikkakrishna, N., K., Hardik, C., Deepika, K. and Sparsha, N. (2019). Short-Term Traffic Prediction Using Sarima and FbPROPHET. 2019 IEEE 16th India Council International Conference (INDICON). DOI: 10.1109/INDICON47234.2019.9028937
Raheem, F. and Iqbal, N. (2021). Forecasting foreign exchange rate: Use of FbProphet. 2021 International Research Conference on Smart Computing and Systems Engineering (SCSE). DOI: 10.1109/SCSE53661.2021.9568284.
Chafiq, T., Ouadoud, M., Elboukhari, K. (2020). Covid-19 forecasting in Morocco using FBprophet Facebook's Framework in Python. International Journal of Advanced Trends in Computer Science and Engineering. 9(5). Retrieved from: http://www.warse.org/IJATCSE/static/pdf/file/ijatcse251952020.pdf https://doi.org/10.30534/ijatcse/2020/251952020
Durairaj, M. and Mohan, K.B.H. (2021). Statistical evaluation and prediction of financial time series using hybrif regression prediction models. International Journal of Intelligent System and Application in Engineering. 9(4). ISSN:2147-6799.
Chakraborty, K., Mehrotra, K., Mohan, C. K., and Ranka, S. (1992). Forecasting the behaviour of multivariate time series using neural networks. Neural Networks,5(6), pp 961-970.
Tanışman, S., Karcıoğlu, A. A., Uğur, A., ve Bulut, H. (2021). Bitcoin fiyatının LSTM ağı ve ARIMA zaman serisi modeli kullanarak tahmini ve karşılaştırılması. Avrupa Bilim ve Teknoloji Dergisi, (32), 514-520.
Süzen, A. A., (2019). LSTM derin sinir ağları ile üniversite giriş sınavındaki matematik soru sayılarının konulara göre tahmini. Engineering science, 14(3): 112-118. Doi:10.12739/NWSA.2019.14.3.1A0436.
Guleryuz, D. ve Ozden, E. (2020). The prediction of Brend Crude Oil Trend Using LSTM and Facebook Prophet. European Journal of Science and Technology, (20), 1-9.
Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles And Practice. OTexts.
Swamidass, P. M. (Ed.). (2000). Encyclopedia Of Production And Manufacturing Management. Springer Science & Business Media.