Forecasting Indonesia’s Composite Stock Price Index with Semiparametric Cubic and Local Gaussian Polynomials

The Composite Stock Price Index (CSPI) serves as a crucial indicator for assessing the performance of the Indonesian capital market, reflecting both economic conditions and investor confidence. Its movements are influenced by macroeconomic factors such as exchange rates, inflation, interest rates, and commodity prices, including oil and gold. Parametric models often fail to capture nonlinear patterns, whereas nonparametric approaches lack efficiency and interpretability. To address this gap, this study develops a semiparametric regression model that integrates a cubic polynomial for parametric effects with local polynomial estimators using Gaussian kernels for nonparametric effects. The results show that the semiparametric model is effective, yielding an MSE of 0.569747, a MAPE of 8.60\%, and an $R^2$ of 85\%. This confirms its ability to capture nonlinear dynamics in the stock market. Moreover, the model provides accurate forecasting and practical insights for investors in portfolio strategies as well as for policymakers in managing financial market stability.