Stock price forecasting by presenting a hybrid model using principal component analysis and rough set theory

In this research, by combining the methods of principal component analysis and the Rough sets, a model is proposed to predict stock prices. First, a number of technical indicators were calculated using the one-year price data of IranKhodro Company. In order to reduce the decision matrix dimension, using the principal component analysis method, new variables were selected so that the maximum characteristics of the initial data were maintained. These variables are used as conditional components in the decision matrix, and the decision variable is next day stock price fluctuation. The data were converted into discrete intervals by different methods and then divided into two groups of learning and control. Then, using the theory of Rough sets on learning data, the decision rules were extracted and their validity on the control data was evaluated. The results obtained from the combined model were compared with the results of the Rough sets model. The advantage of the Principal Components Analysis and Exploratory Factor Analysis methods is the ability to name new factors as the factor of the momentum and the moving average factor, which makes the results more tangible. The percentage of correct predictions of the rules extracted from the hybrid model is higher than the alternative model and the number of rules is lower. In order to verify the reliability of the model, the data of the period of 2002-2017 of IranKhodro Company and also the data of the Iran Saderat bank were studied. The results were consistent with the previous findings.