Time Series Analysis for Number of Monthly Unemployment in Malaysia

Abstract

This study explores challenges in forecasting Malaysia's unemployment rate, focusing on rising trends in recent years. It aims to develop models known as Box-Jenkins (ARIMA), Holt’s Linear, and Naïve methods were used, the best model was identified by comparing Mean Absolute Error, Mean Absolute Percentage Error, and Root Mean Square Error values, and unemployment rates were forecasted with the most accurate model. The dataset included 97 observations was collected from 2016 to 2023 based on the Labor Force Survey (LFS). Statistical analysis revealed that the Naïve method performed poorly, Holt’s Linear method suited linear trends, and the Box-Jenkins (ARIMA) model was the most reliable. During the training phase, the Box-Jenkins method showed the best performance, with the lowest MAE (6.7405), MAPE (0.1950), and RMSE (8.2968), indicating strong fit to the data. In the testing phase, the Naïve Method achieved the lowest errors, with MAE (0.6750), MAPE (0.0227), and RMSE (0.8026). However, the Box-Jenkins method followed with slightly higher errors, still showing good generalization. Therefore, the Box-Jenkins method will be chosen as the best method because it was more compatible to forecast for the long time period compared to Naïve method that only useful to forecast short time period. The use of larger datasets was recommended to improve forecasting accuracy, providing insights for policymakers and researchers in addressing unemployment challenges.