The process of selecting a specialization during university admission represents a critical decision in an applicant’s academic trajectory, directly influencing future career prospects and professional development. However, due to the constraints of limited information available at the time of selection, applicants are susceptible to suboptimal decision-making, which may result in academic dissatisfaction, diminished performance and potential changes in their field of study. In recent years, machine learning methodologies have been extensively investigated for their potential in developing intelligent recommendation systems capable of integrating a broad spectrum of factors to enhance decision-making personalization. This study presents the development of an advanced recommendation system leveraging a neural network-based framework, wherein the core innovation lies in the implementation of an adaptive activation function. Unlike conventional activation functions, this approach dynamically adjusts the response characteristics of neurons, thereby augmenting model flexibility during training and improving predictive accuracy. A comparative evaluation between a baseline model and the proposed adaptive activation model is conducted, demonstrating a statistically significant improvement in classification accuracy for specialization recommendation. The empirical results substantiate the efficacy of the proposed approach in optimizing applicant decision support within the university admission process.