An association analysis and assessment model of college students’ entrepreneurial ability based on graph neural network

Assessing entrepreneurial ability in college students requires analysing a range of interconnected factors, such as psychological traits, academic performance, and social behaviours. Traditional evaluation methods often overlook these complex interdependencies and lack adaptability to diverse student profiles. The research aims to develop an intelligent and interpretable assessment model, Elitist Genetic Optimized Knowledge-Enhanced Graph Attention Network (EG-KE-GAT), for college students’ entrepreneurial ability. Data sources include psychological assessments, academic records, and social interaction data. The dataset is pre-processed through cleaning and normalization, followed by feature extraction using Term Frequency–Inverse Document Frequency (TF-IDF) to capture informative attributes. A KE-GAT is proposed to model the associations among students by embedding domain knowledge into graph-based learning. The network architecture is further optimized using the EG Algorithm (EGA) to enhance classification accuracy and model generalization. Key influencing features such as leadership tendencies, academic engagement, and peer influence were effectively captured. A paired t-test was used to statistically validate the performance differences between the proposed EG-KE-GAT model and baseline models. The EGA optimization led to improved convergence and reduced overfitting. The KE-GAT model outperformed baseline models in accurately assessing entrepreneurial ability levels. The proposed technique had the greatest performance in terms of recall (96.6%), precision (94.7%), accuracy (98.21%), F1-score (97.3%), MAP (0.962), coverage (0.96), novelty (0.82), RMSE (0.25), and MAPE (4.2). Python is used to implement the entrepreneurial ability performance. The proposed EG-KE-GAT framework provides a robust and insightful mechanism for entrepreneurial ability assessment. It enables data-driven decision-making for early identification and development of student entrepreneurial talent.