Comparative Study of Numerical Methods in Multiple Linear Regression For Stock Prediction Jakarta Islamic Index (JII)

Abstract

Stock index predictions in various countries around the world are one of the few challenging issues to be solved. The existence of stock index into a picture of a state of the market in a country, including in Indonesia. Because of the important stock index in a country, it is necessary to predict the future value of the stock. In this research, Jakarta Islamic Index (JII) is used to predict the stock index. To predict the stock indeks value, we proposed a Multiple Linear Regression as a method with coefficient determination using several numerical methods, namely Gauss-Jordan method, Gaussian elimination, and Cramer's rule. Several numerical methods used to solve the system of linear equations on multiple linear regression aim to find out the best numerical method to use. Mean Absolute Percentage Error (MAPE) is used as a comparison of the linear equations system solution method that has been used to test the accuracy of the three methods, and the smaller MAPE value then the prediction models perform better. From the test results it can be concluded that the Gauss Elimination and Cramer's rule method produces the minimum MAPE error value of 0.43% and 0.44%, while for Gauss-Jordan produces MAPE 0.83%.