Comparing 15 Machine Learning Classifiers for University Admissions: Why Random Forest Wins
- ▸Comparing 15 machine learning classifiers on real admissions data, the Random Forest algorithm ranked No. 1, correctly predicting admitted vs. not-admitted status 95.5% of the time with a near-perfect 0.99 ROC-AUC score.
- ▸CGPA (30%) and GRE Score (22%) were by far the most influential factors in the model's predictions — together outweighing softer inputs like statement of purpose, recommendation letters, and research experience.
- ▸The weakest performer was a linear-kernel Support Vector Machine at just 60.8% accuracy, only modestly ahead of a no-skill baseline — underscoring how much tree-based ensemble methods outperform simpler linear models on this kind of data.
If you wanted to predict, from a stack of applications, which students a university would admit, which method gets it right most often? This study tested 15 machine learning approaches against exactly that question, using real admissions data — GRE score, TOEFL score, university rating, statement of purpose, letters of recommendation, CGPA, and research experience — for 400 applicants, expanded to 800 through a statistical resampling technique.
The clear winner was Random Forest, an ensemble method that builds hundreds of decision trees and lets them vote on the outcome. It correctly predicted admission status 95.5% of the time on data it had never seen, with a near-flawless 0.99 ROC-AUC score, a measure of how well the model tells admitted and rejected applicants apart. A close cousin, the Extra Trees classifier, performed almost identically. Both comfortably beat traditional approaches like logistic regression and Naive Bayes, and vastly outperformed a linear-kernel support vector machine, which struggled to break 61% accuracy — barely better than guessing based on the class split alone.
The study's second finding is arguably more useful than the leaderboard itself: which factors actually drive the prediction. CGPA mattered most by a wide margin, followed by GRE and TOEFL scores. Softer factors — the statement of purpose, recommendation letters, and research experience — carried far less weight.
The seven features driving this model — GRE score, TOEFL score, statement of purpose, letters of recommendation, CGPA, and research experience — are precisely the inputs Indian students submitting graduate-school applications to US and other English-medium universities are evaluated on every admissions cycle, making this applicant pool a direct, real-world analogue of the dataset studied here rather than a hypothetical one. India also sends one of the largest cohorts of GRE/TOEFL test-takers globally each year, so a finding that CGPA and standardized test scores dominate softer factors like recommendation letters has direct relevance for how Indian applicants and the coaching/ed-tech industry serving them prioritize preparation effort. Indian higher-education institutions and ed-tech platforms have also begun piloting similar ML-based applicant screening tools domestically, for which this kind of comparative classifier benchmarking offers a useful methodological reference point.
Primary Sources
Cite This Article
EconoLens Research Desk. (2026, August 17). Comparing 15 Machine Learning Classifiers for University Admissions: Why Random Forest Wins. EconoLens. https://www.econolens.co.in/news/ml-classifiers-university-admissions-random-forest
The EconoLens Research Desk reviews academic papers in economics and econometrics, translating cutting-edge research into accessible analysis. Full credit is given to original authors in every review.