I built this project to predict graduate admission chance from academic and applicant-profile features. It compares several regression models and reports model performance using mean squared error and R-squared. I treat the project as a compact supervised learning exercise on a clean tabular dataset.
I used graduate admission prediction as a useful benchmark for regression modeling and feature interpretation. I use the project to demonstrate exploratory analysis, model comparison, and the importance of evaluating multiple baselines rather than relying on one algorithm.
- Source: Public graduate admission prediction dataset used in the notebook.
- File:
data/admission_data.csv - Size: 500 records, based on notebook output.
- Target variable: admission chance.
- Important features: GRE score, TOEFL score, university rating, SOP, LOR, CGPA, and research experience.
- Known limitations: Admission decisions are complex and context-dependent; this dataset is simplified and should not be treated as a real admissions model.
- Loaded and inspected admission data.
- Performed exploratory data analysis.
- Trained and compared multiple regression models.
- Evaluated models with MSE and R-squared.
My notebook reports the following model comparison:
| Model | MSE | R-squared |
|---|---|---|
| Linear Regression | 0.003705 | 0.818843 |
| Random Forest | 0.004346 | 0.787504 |
| Gradient Boosting | 0.004462 | 0.781815 |
| Neural Net (MLP) | 0.028744 | -0.405567 |
| Ridge Regression | 0.003722 | 0.817979 |
| Lasso Regression | 0.003807 | 0.813824 |
| Polynomial Regression (degree 2) | 0.003548 | 0.826512 |
- Polynomial regression performed best in the reported comparison.
- Linear and regularized linear models were competitive.
- The MLP result was poor in this setup, likely due to configuration, data size, or scaling choices.
- The dataset is small enough that validation strategy matters.
- The dataset is simplified and may not represent real admissions decisions.
- I do not use this project to prove causal importance of applicant features.
- Cross-validation and uncertainty estimates should be added.
- Ethical limitations around admissions prediction should be stated clearly.
- Add cross-validation.
- Add residual analysis and feature coefficient interpretation.
- Document the dataset source.
- Add an ethics note about admissions modeling.
git clone https://github.com/BobbY-24/Graduate-Admission-Prediction-Project.git
cd Graduate-Admission-Prediction-Project
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
jupyter notebook notebooks/graduate_admission_prediction.ipynb