期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2019
卷号：97
期号：15
页码：4175-4187
出版社：Journal of Theoretical and Applied
摘要：Stock market forecasting of price/index has always been an important financial subject. Knowing the close price/index based on previous information is useful for investors who need to buy or sell the stock. Most of the applications are focused on building systems with less error and more accuracy. Most traders have used technical analysis tools to predict future stock market movements. Popular methods to find dynamic relationship between input and target output were artificial neural networks that proved to be effective recently. The evolutionary algorithms improve performance in predicting financial market results. This study uses the following: ten technical indicators as inputs, genetic algorithm (GA) to select significant features, backpropagation neural (BPN) to predict future stock price based on features of the previous day and self-organizing map (SOM) to reduce data size. Also, this paper compares three fusion hybrid prediction models, SOM-GA-BPN, SOM-BPN and GA-BPN, with a single model, which is the BPN. Three indices (S&P 500, IBM and NASDAQ) are used in order to evaluate the performance of the proposed hybrid methods. We compare these models with other models such as Support Vector Regression (SVR), Artificial Neural Network (ANN), Random Forest (RF), SVR�ANN, SVR�RF and SVR�SVR fusion prediction models using evaluation measures. The comparison proves the effectiveness and accuracy of the proposed technical indicators and methods.
