A New Engineering Approach to Forecast Tehran Securities Exchange Indices Volatility

Document Type : Research Paper

Authors

Abstract

Journal of Accounting Advances (J.A.A)
Vol. 7, No. 2, 2015, Ser. 69/3
 
 
Extended Abstract
 
A New Engineering Approach to Forecast Tehran Securities Exchange Indices Volatility
 
Dr. Mahdi Moradi*                        Dr. Hadi SadoghiYazdi**        
 Javad Abdollahian***       
 
Introduction
 Forecasting daily securities exchange indices volatility is important in prediction of time series because volatilities in stock market are essentially dynamic, non-linear, complex, nonparametric and amorphous in nature (Kara and et al., 2011; Manish and Thenmozhi, 2005; Abu-Mostafa and Atiya, 1996). According to the risk of stock market securities in recent years, researchers in their studies introduce different models and algorithms for estimating and forecasting securities exchange indices volatility in order to help investors. However, none of these studies have provided non-linear online algorithms to predict indices volatility. Applying engineering approaches, investors can be protected against potential market risks and opportunities for speculators and arbitrageurs to make profit by trading in stock index. Clearly, being able to accurately forecast the stock market index has profound implications and significance for both researchers and practitioners. Therefore, in this study, we use a new engineering approach that accurately predicts stock market indices volatility.
According to many factors that typically cause changes in economic phenomena, nonlinear online models have an important place in the economic literature of prediction models, and already they play an important role in modeling economic relations which are the key step in the short-term and long-term forecasts. (Moshiri and Morovat 1385). In this article, we evaluate the above hypothesis, and in particular to answer the question whether the anticipated volatility of Tehran securities stock indices using KLMS non-linear online algorithm is more accurate than the prediction of the Neural network algorithm based on the LMS?
 
Research Hypothesis
Hypothesis 1: prediction volatility of KOL Index in Tehran securities exchange is more accurate than the prediction of the Neural Network algorithm based on the LMS by using KLMS algorithm.
Hypothesis 2: prediction volatility of price and cash return index in Tehran securities exchange is more accurate than the prediction of the Neural Network algorithm based on the LMS by using KLMS algorithm.  
Hypothesis 3: Prediction volatility of industry index in Tehran securities exchange is more accurate than the prediction of the Neural Network algorithm based on the LMS by using KLMS algorithm.
Hypothesis 4: Prediction volatility of Tehran securities exchange first market index is more accurate than the prediction of the Neural Network algorithm based on the LMS by using KLMS algorithm.
Hypothesis 5: Prediction volatility of fifty top Companies index in Tehran securities exchange is more accurate than the prediction of the neural network algorithm based on the LMS by using KLMS algorithm.
Hypothesis 6: Prediction volatility of financial intermediation index in Tehran securities exchange is more accurate than the prediction of the neural network algorithm based on the LMS by using KLMS algorithm.
Hypothesis 7: Prediction volatility of Free Float Index in Tehran securities exchange is more accurate than the prediction of the Neural Network algorithm based on the LMS by using KLMS algorithm.
 
Research method
Recent studies have shown that there is a nonlinear correlation in stock market indices. So in this study, we used the Kernel least mean square (KLMS) algorithm for prediction volatility of Tehran securities exchange indices, and compared its performance with the neural network algorithm based on the LMS in the short, medium and long term horizons. Prediction performance and reliability of the proposed algorithms are measured using Mean Square Error (MSE), Root Mean Square Error (RMSE), Normalized Mean Square Error (NMSE), Mean absolute difference (MAD), Directional Symmetry (DS), Correct up trend (CP), Correct down trend (CD).
 
Results
Based on the prediction volatility of Tehran securities exchange indices in short term time horizon (prediction volatility of next day), medium term (prediction volatility of next month) and long term (prediction volatility of next year) by using KLMS algorithm and neural network algorithm based on the LMS, it is concluded that the KLMS algorithm performances based on MSE, RMSE, NMSE and MAD in all indices are significantly less than the LMS algorithm. These results show that there is the lower deviation between the actual and predicted values ​​using the proposed KLMS algorithm. Also DS, CP and CD values ​​of KLMS algorithms in short term, medium term and long term horizons ​​are higher than neural network algorithm based on the LMS. This indicates that the appropriate prediction benchmark index is offered.
 
Discussion and Conclusion
Comparing results of KLMS and neural network algorithm based on the LMS based on MSE, RMSE, NMSE, MAD, DS, CP, CD show that the performance prediction made ​​by KLMS is much more accurate than neural network algorithm based on the LMS. Because KLMS algorithm is nonlinear and online algorithm, while neural network algorithm based on the LMS is linear. In the online scenario where data is constantly changing, KLMS network size will grow steadily and implementation challenges will appear. Therefore, the proposed KLMS algorithm provides a good prediction of forecast error and forecast accuracy and can effectively detect and eliminate the noise in financial time series data and improve prediction performance. Overall, the results indicate the predictive power of the KLMS algorithm. 
  

  

 
  

 

* Associate Professor of Accounting, Ferdowsi University of Mashhad
    Corresponding author: Mhd_Moradi@um.ac.ir


** Associate Professor of Computer, Engineering Department, Ferdowsi University of Mashhad


*** MSc of Accounting, Islamic Azad University, Mashhad Branch

Keywords