Forecasting the Stock Price of PT Unilever Indonesia Using the ARCH-GARCH Model with the Application of Kalman Filter

Abstract

PT Unilever Indonesia experiences significant stock price volatility driven by both internal and external factors. This volatility underscores the need for accurate forecasting methods to support investment decision-making and risk management. This study aims to forecast the company’s stock prices using ARCH-GARCH models, enhanced with the Kalman Filter to improve predictive performance. Daily historical stock price data were obtained from the yfinance library. The research methodology consists of several stages, including literature review, data collection, exploratory data analysis (EDA), data preprocessing, forecast modelling, and evaluation. Among the evaluated models, the GARCH(1,2) with a skewed Student’s t error distribution was identified as the best-fitting model, achieving an AIC value of -5.476981. The initial forecast using the GARCH model produced a MAPE of 49.47%, RMSE of 45.56%, and MAE of 37.16%. After applying the Kalman Filter, the model’s forecasting performance improved substantially, with MAPE decreasing to 6.04%, RMSE to 6.01%, and MAE to 5.02%. These results demonstrate the effectiveness of the Kalman Filter in reducing noise, dynamically updating predictions, and enhancing the model’s responsiveness to market fluctuations.