Construction of Intelligent Evaluation Model for Music Aesthetic Education Effect in Colleges and Universities—Application Exploration Based on Informer Time Series Algorithm

Accurate evaluation of music aesthetic education requires effective modeling of long-term temporal dependencies and short-term dynamic variations in heterogeneous educational data. This study proposes an intelligent evaluation framework based on a parallel Informer and BiGRU-Global Attention architecture to analyze multi-source time series collected from questionnaires, classroom interactions, and practical assessments. The Informer module captures global evolutionary patterns through ProbSparse attention and multi-scale decomposition, while the BiGRU-Global Attention branch enhances sensitivity to local fluctuations and key educational events. Experimental validation on 40-week records from 1,984 students demonstrates that the proposed model achieves superior predictive performance with an MAE of 2.3, RMSE of 3.0, and MAPE of 3.5%, while maintaining an inference time of only 8.4 ms. Comparative experiments and statistical significance tests confirm its robustness and effectiveness in dynamic educational assessment. Beyond music aesthetic education, the proposed multi-source temporal fusion strategy provides a transferable data-driven methodology for intelligent signal analysis and sequential decision-making tasks, offering potential reference for information processing frameworks in electromagnetic sensing and human-centered intelligent systems.