MALAYSIA TECH STOCK FORECASTING USING ARIMA: A COMPARATIVE ANALYSIS WITH KALMAN FILTER NOISE OPTIMIZATION

Abstract

Forecasting stock prices in Malaysia's technology sector is difficult due to market noise. To mimic market behavior, many models have been developed and gradually improved with new algorithms and machine learning techniques. Relying on a single model often has drawbacks, so hybrid models were introduced to improve prediction stability. This study used a time series method, Autoregressive Integrated Moving Average (ARIMA) that has been proven to be reliable in forecasting various types of data. However, the ARIMA model often struggles with handling nonlinear and noisy data; thus, hybrid integration of the Kalman Filter was introduced to help reduce noise. The Kalman Filter method is an algorithm that applies the Bayesian theorem in order to predict and update after each observation. Using the Kalman Filter as a hybrid integration helps capture the underlying patterns of ARIMA residuals. As ARIMA creates a general model, the Kalman Filter acts as an extension that analyzes and interacts with the unobservable pattern in the general model. This research used ten years of daily data from 2015 to 2025 from Inari Amertron (INAR), Frontken Corporation (FRONTKN), and ViTrox Corporation (VITROX); the models are evaluated based on Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). The outcome of this study showed that adding a Kalman Filter improves prediction accuracy by lowering noise and stabilizing forecasts. Moreover, this study highlighted the role of Kalman filtering in improving stock price prediction for emerging markets, such as Malaysia.