Abstract
The development of interdisciplinary competence among university faculty increasingly serves as a key driver for educational innovation and knowledge integration. This study proposes a modeling framework based on Artificial Neural Networks for competence identification and pathway optimization. A Multilayer Perceptron (MLP) model is employed to conduct quantitative modeling and nonlinear prediction of 14 influencing factors. The empirical results show that the mean squared error of MLP on the test set is 0.276, which is 4.5% lower than that of One-dimensional Convolutional Neural Network (1D-CNN), and the mean absolute error is reduced by 2.11%. The goodness of fit (R(2)) of MLP reaches 0.843, which is 0.96% and 24.24% higher than that of 1D-CNN and the decision tree, respectively. In multi-class classification tasks, the model attains accuracy rates of 84.09% and 82.35% for the "high" and "low" competence levels, with an overall average accuracy of 79.34%. Normalized weight analysis reveals that cognitive flexibility exhibits the strongest marginal effect within the low-competence group, while at higher levels, competence evolution is jointly driven by external resources and internal motivation. The study further recommends a parallel advancement strategy through cognitive activation, curriculum integration, and institutional empowerment to achieve systemic transformation in interdisciplinary competence development. This study establishes an integrated pathway encompassing indicator identification, predictive modeling, and strategic feedback, providing methodological support and empirical evidence for the precise design of interdisciplinary faculty development mechanisms in higher education.