Forecasting Garanti Bank (GARAN) Stock Price in BIST30 using Artificial Neural Network
Abstract: Today, artificial neural networks are popularly applied to many financial problems such as stock market index estimation, bankruptcy prediction, or corporate bond classification. The studies also focus on the direction of daily index change as much as the stock index value estimate. In some applications it is stated that artificial neural networks restrict data patterns without learning. Artificial neural networks can deliver unparalleled performance that is inconsistent with the complex financial data cause, while offering outstanding learning ability. In addition, data is sometimes so voluminous that learning patterns do not work. The elimination of unnecessary features and the reduction of the size of the data due to the presence of continuous data and large recordings shortens the processing time of the algorithm and gives more generalizable results.
In addition to models such as linear and nonlinear regression analysis, Random Walk, GARCH, and ARIMA, which are traditional estimation techniques, Artificial Neural Networks have started to find use of this field in recent years and have produced more successful results than conventional techniques.
In this study, prices and price directions of a stock of BİST30 index namely GARAN which is the most traded stock in market is estimated with the Feed forward Artificial Neural Networks with the success of 52% in direction. After that a simulation model was created using a trading strategy that calculates how much the ANN model performed in the first 6 months of 2016.