The Fall 2025 project in collaboration with the Massachusetts Department of Economic Research extends last semester’s progress developing an in-house forecasting model for Massachusetts unemployment rate. There are three new focuses for this semester: (1) expanding the feature set to include more alternative data indicators, (2) exploring more innovative model architectures for unemployment forecasting, and (3) gauging the predictive power of sentiment analysis on 10-Q corporate reports for labor market prediction. Thus far the team has put together a comprehensive feature set that incorporates both national and state-level macroeconomic indicators, Google Trends search indices, and 10-Q sentiment. Initial findings show that labor adjacent Google Trends indices are highly predictive of employment outcomes, and Random Forest models outperform traditional autoregressive methods for longer-term prediction horizons.