Abstract
The agro-industrial complex of Kazakhstan faces increasing pressure to improve the efficiency of marketing activities under conditions of limited budgets, regional heterogeneity, seasonal demand, and uneven digital infrastructure. Traditional marketing planning methods often rely on expert judgment or single-objective optimization and therefore cannot adequately address the conflicting objectives of maximizing marketing effectiveness, expanding audience coverage, and minimizing campaign expenditure. This study proposes a hybrid decision-support framework that combines the Non-dominated Sorting Genetic Algorithm II (NSGA-II) with machine learning-based clustering for multi-objective marketing budget optimization in the agro-industrial sector. The model considers five major promotion channels: digital media, television, radio, print media, and events. NSGA-II is used to generate Pareto-optimal budget allocation strategies, while K-means clustering classifies the resulting solutions into interpretable strategy groups. The framework was evaluated using synthetic data and enterprise-level data reflecting marketing conditions in Kazakhstan. Results show that strategies combining digital media and television achieve the highest marketing effectiveness and audience coverage, although they require greater campaign expenditure. Radio and print media remain relevant for cost-sensitive strategies, particularly for enterprises with limited resources. Cluster analysis identified three main strategy groups: cost-efficient, balanced, and high-performance configurations. The findings confirm that marketing budget allocation in the agro-industrial sector should be treated as a multi-objective optimization problem rather than a single-criterion decision. The proposed framework provides both computational optimization and managerial interpretability, supporting data-driven, adaptive, and resource-aware marketing planning for agro-industrial enterprises in Kazakhstan.