A Hybrid Machine Learning Approach for Graduate Admission Prediction and Combined University-Program Recommendation

Researchers Melina Heidari Far and Elham Tabrizi have developed a novel hybrid machine learning framework that tackles the competitive graduate admissions process with impressive precision. Their model, detailed in the arXiv paper "A Hybrid Machine Learning Approach for Graduate Admission Prediction and Combined University-Program Recommendation," achieves 87% accuracy on its test set. It was trained on a unique dataset of 13,000 self-reported application records from GradCafe (2021-2025), which the authors enriched with external data from the OpenAlex API, QS World University Rankings, and Wikidata.

The core innovation is a two-part system. First, a predictive model combines the gradient boosting power of XGBoost with a residual refinement module using k-nearest neighbors to classify admission outcomes. Second, and crucially for applicants, a recommendation module activates for predicted rejections. This system analyzes the applicant's profile against the enriched dataset to suggest viable alternative university and program combinations, reportedly improving the expected probability of acceptance by 70%. The research underscores that university quality metrics are a dominant factor in decisions for competitive pools.