COMPARISON OF LEAST SQUARE SPLINE AND ARIMA MODELS FOR PREDICTING INDONESIA COMPOSITE INDEX

The Indonesian Composite Index (ICI) reflects Indonesia's economic growth. ICI predictions are significant considerations for investors when making investment decisions. Two approaches can be used to predict ICI: parametric and nonparametric approaches. Therefore, this study compares parametric and nonparametric approaches to predict ICI. In its application, the parametric approach requires several assumptions to be met, such as linearity. This differs from analysis with a nonparametric approach that does not require certain assumptions. The parametric approach in this study uses the ARIMA model. ARIMA is widely used to predict time series data. In the nonparametric approach, in this study, we used nonparametric regression based on the least square spline. Spline is chosen because it can handle data that tends to fluctuate by placing knot points when data changes occur. In this study, ICI monthly data obtained from the website investing.com was used. Investing.com is a website that financial analysts often use as a data source to monitor world economic conditions, including the ICI. The Mean Absolute Percentage Error (MAPE) value is determined to assess the accuracy of the prediction. The study results indicate that the analysis with ARIMA cannot meet the assumptions, so ARIMA modeling cannot be continued. Different results were obtained in nonparametric regression modeling based on the least square spline estimator. Prediction of ICI using nonparametric regression based on the least square spline estimator has a MAPE value of 2.613% (less than 10%), which means the model is a highly accurate prediction, meaning modeling using nonparametric regression based on the least square spline estimator is better than the ARIMA model for predicting ICI.