Akademik Veri Yönetim Sistemi
CIENCIA RURAL, cilt.56, sa.4, ss.1-16, 2026 (SCI-Expanded, Scopus)
Accurate forecasting of the wheat producer price index is crucial for economic planning, trade policy, and food security management. Traditional econometric and machine learning approaches often struggle to capture the complex interplay of economic, climatic, and trade-related drivers of price dynamics. This study proposed a novel hybrid framework that integrates graph neural networks, convolutional neural networks, and random forest to enhance predictive performance. The model combines economic indicators, meteorological variables, governance measures, and graph-based trade representations to capture both spatial and temporal dependencies. Data were obtained from the Food and Agriculture Organization for economic and trade indicators across 109 countries, from the Turkish State Meteorological Service for national weather parameters, and from the World Bank’s Worldwide Governance Indicators, specifically the Political Stability and Absence of Violence/Terrorism index, to reflect geopolitical risk. The dataset spans 2000-2023 and was randomly split into 80% training and 20% evaluation subsets.We benchmarked the proposed approach against baseline models. Results demonstrated that the GNN+CNN+RF hybrid, enriched with the Political Stability Index, achieved the best performance, reducing Mean Squared error to 158.44 and Root Mean Squared Error to 12.59, while attaining the highest R² score of 0.962. These findings highlighted the advantages of integrating graph-based learning with deep learning and ensemble methods, further strengthened by governance-related risk indicators.