With the advancement of educational informatization and personalized learning, scientific evaluation of learning outcomes has become crucial for educational quality assurance. This paper proposes a hybrid evaluation approach integrating grey correlation analysis, BP neural network, and fuzzy evaluation based on EEG data for assessing personalized learning effects. The method establishes objective evaluation indicators through EEG data analysis, enabling real-time monitoring and assessment of the learning process. By adopting a multi-model fusion strategy, the accuracy and reliability of the evaluation are enhanced. The evaluation framework encompasses data preprocessing, feature extraction, model fusion, and result validation. Empirical research in primary education demonstrates that this method achieves 89% consistency with expert evaluation, 85% accuracy in cross-validation, and a correlation coefficient of 0.82 with academic performance. Over an eight-week intervention period, students showed significant improvements: attention levels increased by 35%, learning efficiency improved by 40%, and assignment quality enhanced by 28%. The research findings provide a new paradigm for data-driven educational evaluation and make significant contributions to advancing scientific and personalized development in educational assessment.