<![CDATA[Stocks are the investment that is much in demand by investors because they are able to provide a high level of profit with a certain risk. Therefore, stock price forecasting is very important to maximize investment returns. The purpose of this study was to forecast stock prices using the support vector regression (SVR) method by utilizing linear, RBF, sigmoid, and polynomial kernel functions. Parameter optimization is carried out using a grid search algorithm that applies the concept of cross validation. After training and testing the model, the best SVR model is obtained using a polynomial kernel with parameters , , and , which produces an accuracy of 0,99211, RMSE of 0,01027, and MAE of 0,00723 on the training data and produces an accuracy value of 0,99389, RMSE of 0,01988, MAE of 0,01323, and MAPE of 0,02709 on data testing. Forecasting results for the next 85 periods using the best SVR model have a MAPE of 6,45%, this means that the SVR model obtained is able to predict closing stock prices much better than the ARIMA model which has a MAPE of 20,68%.]]>
Copyrights © 2024
