GradMit
Graduate Admission Prediction Platform
Built and deployed a Logistic Regression admission prediction platform achieving 91.2% accuracy and 0.975 ROC-AUC on unseen data.
System Architecture
Prediction path and evaluation evidence from the GradMit repository.
Applicant profile
Academic and profile attributes are captured through a structured prediction form.
Feature preprocessing
The submitted values are validated and transformed to match the trained model schema.
Logistic Regression
The classifier estimates an admission outcome and probability from the prepared features.
Held-out evaluation
Performance is reported on unseen data with 91.2% accuracy and 0.975 ROC-AUC.
Flask web interface
The result page communicates the prediction and supporting evaluation context.
A deployed admission prediction workflow evaluated on 500 records.
Project Overview
GradMit is a machine learning-based admission prediction platform designed to estimate graduate admission outcomes using applicant academic and profile information. The project aims to help prospective students assess their admission prospects and understand the factors influencing admission decisions.
The system processes applicant information such as academic performance, standardized test scores, and profile-related attributes through a machine learning pipeline.
A Logistic Regression classifier was trained and evaluated with accuracy, ROC-AUC, a confusion matrix, and feature-contribution views.
The final model achieved 91.2% accuracy and 0.975 ROC-AUC on unseen data and was integrated into an interactive web application for instant admission predictions.
The platform combines data preprocessing, feature engineering, model evaluation, and deployment into a complete end-to-end machine learning workflow.
Problem
Graduate applicants need a fast way to estimate admission likelihood and understand how profile-based factors influence outcomes.
Solution
Built an end-to-end Logistic Regression workflow that preprocesses applicant features, evaluates held-out performance, and serves predictions through an interactive Flask application.