Artificial Intelligence in Economic Forecasting: Applications in Inflation Prediction and Macroeconomic Analysis

This paper examines how artificial intelligence (AI), particularly statistical machine learning (ML), can complement established econometric approaches to inflation forecasting and macroeconomic analysis. Inflation is a difficult prediction target because it is measured with delay, revised in some datasets, driven by multiple interacting forces, and subject to structural change. Linear time-series and structural models remain valuable because they encode persistence, economic mechanisms, and interpretable scenarios. ML methods can add value where information sets are wide, relationships are nonlinear, interactions matter, or data arrive in mixed formats. The central question is therefore not whether Al replaces econometrics, but when and under what validation conditions it improves a policy-relevant forecast. A literature-based synthesis reviews foundational inflation forecasting research, contemporary ML studies, central-bank applications, and institutional data sources. The evidence is conditional. Data-rich ML models improved on selected benchmarks in a prominent U.S. application, yet later replications found that relative performance changes across countries, benchmarks, methods, and episodes. An ECB quantile-regression-forest application found small advantages for core inflation but mixed results for headline inflation. Text embeddings of central-bank communication have also shown forecast information in one euro-area application, although this remains specialized evidence. Broader IMF and OECD studies on output growth and GDP nowcasting are discussed as adjacent, not direct, inflation evidence.