Predicting Student Admission to University

Project Overview

This project predicts the likelihood of a student being admitted to a university based on key academic and personal attributes. The model assists applicants in assessing their admission chances and provides universities with a data-driven approach to evaluating candidates.

Key Insights

• GRE, TOEFL, CGPA, and research experience strongly influence admission probability.
• University rating, SOP, and LOR have a moderate impact on admission chances.
• Logistic Regression was chosen for its efficiency, interpretability, and suitability for binary classification.
• Feature scaling (StandardScaler) improved model performance and ensured consistency.
• Achieved 94% accuracy, with an F1-score of 91%, ensuring a balance between precision and recall.
• ROC AUC score of 0.94, demonstrating strong model performance in distinguishing admission outcomes.
• Deployed an interactive Streamlit web app for real-time user predictions.

Technical Implementation

• Data Preprocessing:
  • Checked for missing values and handled inconsistencies.
  • Scaled numerical features using StandardScaler for better model performance.
• Model Selection:
  • Implemented Logistic Regression, a reliable algorithm for binary classification.
  • Used GridSearchCV for hyperparameter tuning to optimize performance.
• Model Evaluation:
  • Measured accuracy, precision, recall, and F1-score for performance assessment.
  • Generated a confusion matrix and ROC curve to evaluate classification performance.
• Deployment:
  • Developed a Streamlit web app for real-time predictions.
  • Packaged the model using Pickle for efficient loading and inference.

Live Preview

Video Preview

Key Learnings

• Feature engineering is crucial: Selecting and scaling the right features significantly improves model performance.
• Logistic Regression works well for binary classification with interpretable results.
• Model evaluation should go beyond accuracy: Precision, recall, and F1-score provide deeper insights into performance.
• Pipeline automation enhances efficiency and makes the workflow more reproducible.
• Deploying ML models using Streamlit allows users to interact with predictions easily.
• Real-world applications require continuous improvement, such as expanding the dataset or testing other models for further enhancement.