JEET : AI Powered JEE and NEET Dropout After Class 12 Prediction app

Overview

JEET is a Machine Learning project designed to predict the likelihood of students dropping out after Class 12, focusing on those preparing for or appearing in the Join Entrance Examination (JEE) and/or National Entrance cum Eligibility Test (NEET) in India. BY analysing academic and socio-economic data, JEET helps educators, counselors, and students to identify at-risk indivuiduals for timely intervention.

Structure

JEE_Dropout_ML_Project/
├── .gitignore
├── LICENSE
├── Plan.md
├── README.md
├── 01_Data/
│   ├── 01_raw/
│   │   └── JEE_Dropout_After_Class_12.csv
│   ├── 02_Cleaned and Engineered/
│   │   ├── Feature_Engineered.csv
│   │   └── JEE_Dropout_Cleaned.csv
│   └── 03_final/
│       └── JEE_Dropout_Final.csv
├── 02_Data Analysis/
│   ├── 01_Exploration.ipynb
│   ├── 02_Cleaning.ipynb
│   ├── 03_Feature_Engineering.ipynb
│   └── 04_Merging_Final.ipynb
├── 03_notebooks/
│   └── Prototypes_models.ipynb
├── 04_src/
│   ├── evaluation.py
│   └── jee_dropout_model.py
├── 05_models/
│   └── Model_JEET.pkl
├── 06_deployment/
│   ├── Home.py
│   ├── requirements.txt
│   ├── model/
│   │   └── JEET.pkl
│   └── pages/
│       └── Data.py
│       └── JEET.py
├── gitignore
├── LICENSE
└── README.md

Features

• Predicts dropout risk using student academic and socio-economic data
• Handles imbalanced data with SMOTE for better accuracy
• User-friendly web app built with Streamlit
• Enables educators and students to take data-driiven actions

Dataset

includes features such as

• Academic scores (Class 10, Class 12, JEE/NEET scores)
• Attendance records
• Socio-economic factors (family income, parent's education)
• Target Labele: dropout status (Yes/No)

Installation

1. Clone the Repository

git clone https://github.com//Chracker24/JEE_Dropout_ML_Project.git
cd JEE_Dropout_ML_Project

2. (Optional) Create a virtual environment

python -m venv venv
source venv/Scripts/activate   #MacOS: venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

Usage

1. Navigate to Home.py in 06_deployment
2. Run the Streamlit application using Bash

streamlit run Home.py

3. Input stuudent details via the web interface to get dropout risk predictions
4. Talk with Google Gemini powered JEET chatbot about your shortcomings and help

Methodology

• Data Cleaning, encoding cateogorical variables, and feature scaling and engineering
• Addressed class imbalance using SMOTE
• Trained Random Forest Classifier as the main model
• Evaluated with accuracy, precision, recall, F1-Score and ROC-AUC
• used SHAP for model interpretability

Results

• Achieved approximately 73% accuracy
• This is due to presence of leaky data that got removed and will be fixed in later iterations with better data
• Nonetheless, Improved recall with data balancing
• Important predictors : socio-economic status, peer pressure and mental health

Future work and iterations

• Incorporate psychological and motivational data
• Explore advanced models (XGBoost, deep learning)
• Add personalized intervention suggestions
• Expand dataset for broader applicability and fix leaky data situation

Tech Stack

• Python, scikit-learn
• imbalanced-learn (SMOTE, Pipeline)
• pandas, numpy, seaborn, matplotlib
• SHAP (Interpretabiility)
• Streamlit (deployment)

Contributors

Christy Chovalloor - Software Engineering Student, Queen's University Belfast Linkedin Github

Documentation

Documentation will be up soon and I will surely notify through my linkedin. Stay Tuned!

License

MIT License

Contact

For questions or collaboration, contact through email : Chr2412@hotmail.com Email me