Abstract
Abstract Accurate forecasting of agricultural prices is essential for ensuring food security, effective market planning, and informed policymaking. However, predicting prices, especially for perishable commodities like tomatoes, is challenging due to seasonality, sharp fluctuations, and interconnected market dynamics. Traditional univariate models often fall short in capturing these complexities. This study addresses these challenges by forecasting tomato prices in three major Indian markets, namely Delhi, Lucknow, and Kanpur, using Multivariate Singular Spectrum Analysis (MSSA). MSSA decomposes multiple related time series into trend and periodic components, enabling the joint modeling of inter-market relationships. The multivariate decomposition significantly enhances forecasting accuracy by capturing shared patterns and dependencies across markets. We compared the performance of MSSA with three benchmark models: Vector Auto Regressive (VAR), Singular Spectrum Analysis (SSA), and Auto Regressive Integrated Moving Average (ARIMA), across short, medium, and long-term horizons (3, 6, and 12 months). The results show that MSSA consistently outperforms all other forecasting methods, delivering more accurate and robust forecasts. These improved forecasts enable farmers to make timely decisions, help agribusinesses optimize operations, and assist policymakers in designing effective market interventions, ultimately contributing to a more resilient and well-informed agricultural marketing system.