A Hybrid and Reliable Method Integrating Depth and Technical Analysis with Machine Learning Techniques for Predicting Stock Prices

Abstract

It is quite complicated and difficult to predict the price changes direction of stock, because of the price changes of stock are non-linear. Some statistical methods are used to estimate these price changes direction, but these methods are often inadequate in complex stock markets. In this thesis study, a method which tries to predict the changes direction of price in capital markets based on BIST stocks is presented. This method is also the basis for fair and safe trading of stock market transactions with depth analysis. First of all, stocks which have similar financial structure were selected by using fundamental analysis. In the long term, fundamental analysis data were used to find the best yields. Then, the depth and technical analysis data were used to construct the feature vector. These data were obtained from the order books received from Borsa Istanbul. The information in depth analysis was used to estimate the future trend in financial prices. However, in the depth analysis, the fact that the sales order is too high may not mean that the price of the stock will decrease, but the fact that the purchase orders are too high may not mean that the price will increase. It is not the right method to decide whether the stock will rise or fall just by looking at the depth analysis. Therefore, technical analysis was used as a second indicator. In the technical analysis, the historical data and price movements of the stock are analyzed, and the possible future price movements are estimated. In the price iv estimation, backpropagation neural networks and random forest were used. In this thesis, the combined use of fundamental analysis, technical analysis and depth analysis has contributed to studies in this area.