Forecasting closing returns of Borsa Istanbul Index with Markov Chain Process of fuzzy states

Esin Kiral, Berna Uzun
2017 Pressacademia  
Purpose- The estimation regarding to the exact daily price of the stock market index has always been a difficult task in the business sector. Therefore, there are numerous research studies carried out to predict the direction of stock price index movement. Methodology-Classical Markov chain model (MC) is commonly used for this prediction and it gives valuable signals about the movements of the closing returns of the stock market index. In this paper, we propose Markov Chain Model with Fuzzy
more » ... odel with Fuzzy States (MCFS) to predict the closing returns of Borsa Istanbul (BIST 100) index using triangular fuzzy numbers. We apply this method to hold the information while system moves between the extreme values of the states. Findings-With this study, we show that the use of MCFS for the selected period provides a higher forecasting accuracy to the investors compared to MC model. Conclusion-Markov chains of the fuzzy states defines a stochastic system more precisely than the classical Markov chains and it gives more sensitive future prediction opportunities. It can be used for estimating returns of individual common stocks and also for the other investment instruments. Stock Index, which is related to the changes in the stock prices, plays a significant role in the business sector for the performance valuation. Many factors might have an effect on the stock market index such as the political events, general economic conditions, trader's expectations etc. Stock index prediction has been very interesting research topic for many years. Due to it's complex, dynamic and highly non-linear data over time, it is a very difficult task to predict the exact daily price of the stock market index. The direction of the stock market index correlates with the movement of the price index. Estimating the direction significantly influences the decision of the financial traders about buying or selling an instrument. Hence, the stock index prediction can provide investors to gain profit in the stock exchange. This paper is organized as follows: we provide literature rewiew and concept of the paper in Section 2. In Section 3, the data set, proposed methodology which is related to the fuzzy sets and the Markov chain process of the fuzzy sets are given. In Section 4, findings and discussions about the stock return estimation with both MCFS and MC models are presented. Our conclusions are given in Section 5. LITERATURE REVIEW Recently, various estimating models have been presented and applied for the stock market analysis. Hidden Markov Model (Rabiner & Juang, 1993; Rabiner, 1993) , widely implemented estimating models to estimate stock market data. Box and Jenkins (1976) , used the Time series analysis to estimate and control. White ( ,1989 used Neural Networks to estimate stock market of IBM daily stock returns. Henry (1993), used ARIMA model to predict the daily close and morning open price. But these conventional methods are not useful when non linearity exists in time series. Chiang, Urban and Baldridge, (1996) , have estimated the end-of-year net asset value of mutual funds via ANN model. Kim and Han (2000) , showed there are complex dimensionality and buried noise at the stock market data so that makes it hard to reestimate the ANN parameters. Romahi and Shen (2000) showed that, ANN sometimes suffers from over fitting problem. They have evolved rule, which depends on expert system and generated a method to predict financial market behaviour. Later on, hybridization models have been used for estimating financial behaviour. All of these methods has required expert knowledge to deal with the aforementioned problems. Hassan and Nath (2005) , used HMM for optimizing the system in a better way. Hassan, Nath and Kirley (2006) , integrated HMM and fuzzy logic rules to evolve the prediction accuracy on nonstationary stock data sets. Following this, Hassan, Nath and Kirley (2007) , presented a fusion model of HMM, ANN and GA to estimate stock market. Badge (2012) used technical indicators as an input variable instead of stock prices for analysis. Gubta and Dhingra (2012) considered the fractional change in stock value and the intra-day high and low values of the stock to train the continuous HMM.
doi:10.17261/pressacademia.2017.362 fatcat:mmoqs6bzwnc7xla7i5u7ftbv2a