The application of ensemble connectionist models for analyzing sectoral economic indicators is relevant in modern intelligent analytical systems. Artificial neural networks (ANNs) in ensemble configurations improve approximation accuracy and model interpretability while reducing computational limitations. This research seeks to increase the efficiency of analyzing sectoral macroeconomic indicators by developing approximation models grounded in the ensemble-based connectionist approach. The research proposes an ensemble ANN approximation model based on boosting and population-based metaheuristics to enhance scalability and robustness. Neural network models such as the two-layer perceptron, Elman, and Jordan networks are used as base approximators. The use of metaheuristic algorithms enables parallel training and improves performance. The proposed method provides resilience to incomplete or noisy data, simplifies structural and parametric identification, and eliminates the need for prior assumptions regarding factor distribution or correlation. The boosting-based ensemble ensures adaptive error correction and increased generalization ability. The numerical data demonstrate the effectiveness of the suggested model for sectoral macroeconomic analysis (coefficient of determination 0.89). The study employs a dataset of economic indicators sourced from the Enerdata platform. Prospective research avenues include extending the proposed methodology to a wider range of artificial intelligence applications, alongside the advancement of innovative techniques for analyzing economic indicators with improved precision and interpretability.