Comparative analysis of stochastic models and machine learning algorithms for inflation rate prediction in Nigeria

Inflation forecasting is critical to economic planning, particularly in developing economies like Nigeria, where inflation volatility significantly impacts policymaking, investment decisions, and overall economic stability. This study evaluated the predictive performance of traditional stochastic processes such as Vasicek, Cox–Ingersoll–Ross (CIR), and Geometric Brownian Motion (GBM) against three machine learning algorithms: Random Forest, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), in modeling Nigeria’s inflation trends. The analysis was based on “All-items inflation rates” data from January 2003 to December 2024. The study uncovered that whereas stochastic models successfully captured the hypothetical inflationary change, their predictive accuracy was moderately restricted compared to machine learning methods. In particular, the Random Forest model outperforms stochastic approaches in terms of accuracy, robustness, and overall performance across key evaluation metrics. This research advocates for a paradigm shift in Nigeria`s economic modelling strategies by emphasizing the integration of advanced machine learning methods into inflation forecasting.