Stock Price Forecasting Using Autoregressive With Exogenous Variable Support Vector Regression (ARX – SVR)

Stock prices move fluctuate continuously and dynamically at all times, so stock price predictions are needed to maximize profits for investors and avoid losses due to the characteristic of stock prices. Autoregressive (AR) model is a forecasting method and has weaknesses against nonlinear patterns. In addition to using linear modeling, forecasting stock prices can use the Support Vector Regression model which offers a global optimal solution that works with data maps to high-dimensional spaces and has good performance with time series problems. The addition of exogenous variables X to the model can also improve forecasting accuracy. Forecasting will be done using significant lags as input to Support Vector Regression. The modeling results show that the ARX-SVR model with X as an outlier exogenous variables provides the best out-of-sample data forecasting results for the case study of stock closing price forecasting. This model provides forecasting results with Symmetric Mean Absolute Percentage Error (sMAPE) 5.382430%.