Evaluation of machine learning techniques for forecasting Malaysia's consumer price index: a comparative study

Abstract

Ensuring price stability through accurate measurement and management of the Consumer Price Index (CPI) fosters a stable economic environment conducive to sustainable growth, investment, and employment. As a key economic indicator, the CPI provides a comprehensive assessment of inflation, purchasing power, and the cost of living, serving as an essential tool for policymakers, businesses, and consumers. In Malaysia, the CPI has steadily increased, reflecting a stable inflation rate. Recognizing the need for low and stable inflation, governments prioritize this goal to enhance economic prosperity and societal well-being. Accurate CPI forecasting is crucial for economic stability and informed financial decisions. Machine learning (ML) models have demonstrated significant potential for improving CPI forecasting accuracy over traditional methods. However, research specifically targeting CPI and inflation rate forecasting in Malaysia remains limited. This study evaluates the performance of five ML techniques: Autoregressive Integrated Moving Average (ARIMA), Geometric Brownian Motion (GBM), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Adaptive Neuro-Fuzzy Inference System (ANFIS), in predicting Malaysia’s CPI. The models are assessed by comparing their prediction to actual CPI data from October 2022 to September 2023. Results indicate that GRU model performs best, exhibiting the lowest RMSE, MSE, and MAPE scores, thereby highlighting a consistent upward trend in inflation. This study encourages further exploration of Malaysia’s inflation using advanced ML models or hybrid approaches to enhance forecasting accuracy.