Abstract
Investment assessment in the agricultural sector remains a complex problem owing to uncertainty, incomplete information, and the heterogeneous nature of economic and environmental data, particularly in developing economies. This study proposes a multi-agent hybrid model to evaluate the investment attractiveness of Kazakhstan’s agro-industrial complex. The methodology integrates fuzzy logic, artificial neural networks, and optimization algorithms into a modular multi-agent architecture. The model is based on international datasets from the OECD and FAO covering the period 2010–2023, including Producer Support Estimate (PSE), Agri-Environmental Indicators (AEI), Agricultural Outlook projections, and foreign direct investment (FDI) inflows. The data were normalized using the min–max method and transformed into fuzzy representations through triangular membership functions to capture uncertainty and expert judgment. Specialized agents perform data acquisition, preprocessing, factor analyses, and prediction tasks. Neural networks optimized using genetic algorithms and particle swarm optimization were trained on fuzzy-transformed indicators to generate an integrated investment attractiveness index ranging from zero to one. The experimental results demonstrate high predictive performance, with MAE = 0.083, ROC-AUC = 0.87, and F1-score = 0.81, confirming the robustness and stability of the proposed approach. The results indicate that combining fuzzy logic for interpretability, neural networks for predictive accuracy, and multiagent systems for scalability significantly improves investment assessment under uncertainty. The proposed framework can be practically applied by policymakers, investors, and agricultural stakeholders to support transparent, adaptive, and evidence-based investment decision-making in Kazakhstan’s agro-industrial sector.