خلاصة:
The main objective of this study is to find out whether an Artificial Neural Network (ANN) will be useful to predict stock market price, which is highly non-linear and uncertain. Specifically, this study will focus on forecasting TSE Price Index (TEPIX) as the most significant index of Iran Stock Market. Many data have been used as inputs to the network. These data are observations of 2000 days for a period of 9 years from 02/29/2000 to 12/03/2008. Data are divided into two categories; fundamental and technical data. The fundamental data used here are principal economic values like Dollar/Rials Exchange Rate, Gold price and Oil price. The technical data used are technical indices such as Moving Average (MA), Moving Average Convergence/Divergence (MACD), Relative Strength Index (RSI), Rate of Change (ROC), Momentum (MOM) and daily trading volume of stocks. The selected data are divided into training set and test set, in order to be entered into the network and the remaining 10% was used as the testing set. Training set consists 90% of data. This classification uses 3 different approaches to assemble the training and test data, including random, deterministic and consecutive selection. Here, a feed-forward neural network (FFNN) with the most suitable algorithm for finance (i.e. Back Propagation algorithm) was used for the prediction. Predictions were made for the next day of TEPIX with a 3-4-1 topology and 1500 epochs. The performance of the ANN was evaluated by MSE. Finally, the results showed that ANN could properly recognize the relationships between fundamental and technical data and TEPIX, so that the prediction of the next day was quite possible.
ملخص الجهاز:
Keywords: Short-term Prediction, Forecasting, Tehran Price Index (TEPIX), Artificial Neural Network (ANN) JEL classification: G17 1.
Using Artificial Neural Network (ANN) in predicting stock market behavior, as a better technique compared to other statistical techniques which are limited when encountering non-linear data set, this research attempts to predict TSE price index behavior.
So, using ANN for the prediction of stock price index is for its ability to learn non-linear relationships between inputs and outputs based upon data set that is introduced to the network as the training set to be learned and recognized by the network.
2. Literature review The ANN has been used in several areas of finance, such as forecasting bankruptcy and business failure (Tam & Kiang, 1990; Wilson & Sharda, 1994); foreign exchange Stock Exchange Price Index (TEPIX)… rate (Refenes & Zaidi, 1995; Yao & Tan, 2001; Nag and Mitra, 2002); stock prices (White, 1988; Yoon & Swales, 1991; Yao, & Poh, 1995; Yao, Tan, & Poh, 1999; Kohara, et al.
Later researches by Yoon and Swales (1991), Steiner and Wittkemper (1995), Yao and Poh (1995b), and others have proved that the ANN could predict stock price better than statistical techniques.
For stock market prediction, input and output data are presented to the network simultaneously.
5. Proposed Method Here, the purpose of the model is to predict Tehran Stock Exchange Price Index (TEPIX) and its changes using FFNN (ANN) trained with Back Propagation.