Forecasting Vietnam Inflation Using Machine Learning Approaches: A Comprehensive Analysis

Purpose: This study investigates the predictive ability of selected machine learning methods for inflation prediction in Vietnam. Design/methodology/approach: This study computes forecasts using autoregressive integrated moving average, extreme gradient boosting, linear regression, random forest, K-nearest neighbour, four variants of the recurrent neural network, and causal convolutional neural network. This research assesses their properties according to criteria from the optimal forecast literature. Then, their performance is compared with the predictions of the International Monetary Fund and Asian Development Bank used by the State Bank of Vietnam as a policy benchmark tool. Findings: Although there is no single best model to predict inflation for various horizons, the findings suggest that the K-nearest neighbour (KNN) model provides better forecasts than others for the 12-month horizon. These forecasts are relatively in line with the projections of well-known international organisations under several conditions. The KNN forecast even outperformed those when considering the COVID-19 crisis. Research implications: The results suggest that the machine learning models selected in this study could be used as an additional benchmark tool for policy decision-making under uncertainty, offering a data-driven approach to supplement traditional economic judgment. Originality/value: This study is the first attempt to employ different advanced machine learning methods to predict inflation in Vietnam. More importantly, these results are then compared with other conventional ones and benchmark forecasts for robustness checks.