Multi-agent system improves forecasting by weighting diverse evidence
FACTS-MAS: A Multi-Agent Architecture for Adaptive Forecasting from Heterogeneous Evidence
Artificial Intelligence
Summary
Making accurate predictions about things like housing or unemployment can be tricky because different types of information help more or less depending on the situation. The authors created FACTS-MAS, a system where different agents each focus on a type of information—like seasonal trends or special events—and make their own forecasts. The system then figures out which agents are most helpful and combines their predictions while ignoring unhelpful ones. Tests on real housing and unemployment data showed this approach works better than some common forecasting methods.
What this means in practice
- •For urban planners: Use adaptive multi-agent forecasts to better predict housing inventory changes across different cities and time horizons.
- •For economic analysts: Improve unemployment claim forecasting by combining specialized models weighted by historical predictive skill.
Authors
Madumita Karthikeyan, Umang Rajnikant Bid, Rong Pan
Abstract
Real-world forecasting often requires integrating heterogeneous evidence whose predictive value varies across horizons, operating regimes, and domains. We introduce FACTS-MAS, a domain-agnostic multi-agent forecasting framework that decomposes evidence into specialized channels for temporal dynamics, exogenous context, seasonality, structural information, and discrete events. Each agent generates an independently testable forecast through a common interface, while leakage-safe historical validation measures its predictive skill relative to a naive reference. These skill estimates are used to weight and combine agent forecasts, so that uninformative evidence is either removed or downweighted rather than forced to participate. A calibration procedure selects domain-specific constant means based on historical validation folds. We evaluate FACTS-MAS on weekly for-sale housing inventory from Zillow and weekly unemployment insurance claims from FRED, for 15 U.S. metropolitan areas at 4-, 8-, and 13-week forecasting horizons. Our method outperforms both a naive last-value forecast method and the TimesFM foundation model. However, a factor-adjusted exponential smoothing model performs as well as the fused forecast. The evidence therefore supports FACTS-MAS as a systematic protocol for testing, calibrating, and recomposing evidence channels before assembling them into the strongest single model for a specific application domain. Code, data, and results are available at https://github.com/Madumita24/FACTS-MAS.