"Technická a fundamentální analýza pro predikci akciového trhu pomocí dolování dat"

Abstract

Predicting stock prices is an essential objective in the financial world. Forecasting stock returns and their risk represents one of the most critical concerns of market decision makers. This thesis investigates the stock price forecasting with three approaches from the data mining concept and shows how different elements in the stock price can help to enhance the accuracy of our prediction. For this reason, the first and second approaches capture many fundamental indicators from the stocks and implement them as explanatory variables to do stock price classification and forecasting. In the third approach, technical features from the candlestick representation of the share prices are extracted and used to enhance the accuracy of the forecasting. In each approach, different tools and techniques from data mining and machine learning are employed to justify why the forecasting is working. Furthermore, since the idea is to evaluate the potential of features in the stock trend forecasting, therefore we diversify our experiments using both technical and fundamental features. Therefore, in the first approach, a three-stage methodology is developed while in the first step, a comprehensive investigation of all possible features which can be effective on stocks risk and return are identified. Then, in the next stage, risk and return are predicted by applying data mining techniques for the given features. Finally, we develop a hybrid algorithm, based on some filters and function-based clustering; and re-predicted the risk and return of stocks. In the second approach, instead of using single classifiers, a fusion model is proposed based on the use of multiple diverse base classifiers that operate on a common input and a meta-classifier that learns from base classifiers’ outputs to obtain a more precise stock return and risk predictions. A set of diversity methods, including Bagging, Boosting, and AdaBoost, is applied to create diversity in classifier combinations. Moreover, the number and procedure for selecting base classifiers for fusion schemes are determined using a methodology based on dataset clustering and candidate classifiers’ accuracy. Finally, in the third approach, a novel forecasting model for stock markets based on the wrapper ANFIS (Adaptive Neural Fuzzy Inference System) – ICA (Imperialist Competitive Algorithm) and technical analysis of Japanese Candlestick is presented. Two approaches of Raw-based and Signal-based are devised to extract the model’s input variables and buy and sell signals are considered as output variables. To illustrate the methodologies, for the first and second approaches, Tehran Stock Exchange (TSE) data for the period from 2002 to 2012 are applied, while for the third approach, we used General Motors and Dow Jones indexes.