Forecasting Inflation in the United States Using Machine Learning: A Disaggregated Approach by State

This research explores the use of machine learning (ML) models to enhance inflation forecasting in the United States through a disaggregated approach at the state level. Traditional econometric models, such as ARIMA, have long been used for inflation prediction but often fail to capture regional inflation dynamics. Given the increasing complexity of economic environments, this study investigates whether ML techniques, including neural networks and random forests, can provide more accurate forecasts than conventional models. Using inflation data from the Federal Reserve Economic Data (FRED) and state-level economic indicators, this research develops and compares multiple forecasting models. Benchmarks include autoregressive models and the random walk model, which historically perform well in inflation forecasting. Model performance is evaluated using metrics such as Root Mean Squared Error (RMSE). By examining localized inflation trends and key economic drivers, this study contributes to a more precise understanding of inflationary pressures across different regions. The findings have significant implications for policymakers, businesses, and researchers by providing insights into how disaggregated data and advanced ML techniques can improve macroeconomic forecasting and inform more targeted policy interventions.