Smart Student Performance Prediction & Academic Analytics Platform

Project Overview

This project is an end-to-end Machine Learning solution designed to predict student academic performance based on various academic, behavioral, and demographic factors. It features a comprehensive ML pipeline, a professional Streamlit web application, Explainable AI (SHAP) integration, and automated PDF reporting capabilities.

Problem Statement

Educational institutions often struggle to identify at-risk students before final evaluations. Traditional methods of performance tracking are reactive rather than proactive.

Objectives

• Predict final academic performance accurately.
• Identify at-risk students early.
• Understand the key factors affecting student performance.
• Support data-driven interventions.
• Provide explainable predictions to educators and administrators.

Dataset Description

The base dataset is the Student Performance Dataset (originating from Kaggle/UCI), which includes demographics and test scores. It has been augmented with realistic engineered features to better capture student behavior:

• attendance: Percentage of classes attended.
• assignment_score: Average score on continuous assignments.
• internal_assessment: Internal exam performance.
• participation_score: Classroom engagement metric.
• study_hours: Self-reported weekly study hours.

Feature Engineering

Several advanced features are derived to improve prediction capabilities:

• Overall Internal Score: Average of core subject scores.
• Consistency Score: Inverse of standard deviation across subjects.
• Engagement Score: Weighted combination of attendance and participation.
• Academic Risk Index: Risk metric based on attendance, study hours, and assessments.
• Performance Index: General performance metric.
• Study Effectiveness Score: Ratio of predicted performance to study hours.
• Risk Category: Classification into Low, Medium, or High risk.

Exploratory Data Analysis (EDA) Summary

Comprehensive Univariate, Bivariate, and Multivariate analyses are provided in the Jupyter Notebook, featuring:

• Distribution analysis of academic scores and attendance.
• Correlation heatmaps to identify relationships between behavioral metrics and final performance.
• Pairplots and categorical visualizations (boxplots, violin plots).

Models Used

1. Linear Regression
2. Random Forest Regressor
3. XGBoost Regressor

Evaluation Metrics

Models are evaluated based on:

• R² Score (Primary selection criteria)
• Root Mean Squared Error (RMSE) (Secondary criteria)
• Mean Absolute Error (MAE)
• Mean Squared Error (MSE)
• Adjusted R²
• Cross-Validation Mean and Standard Deviation

Explainable AI (SHAP)

Integration of SHAP (SHapley Additive exPlanations) provides interpretability:

• Summary Plots & Feature Importance
• Dependence Plots
• Waterfall Plots for local explanations

(Note: Streamlit features, PDF generation, and deployment instructions are part of subsequent phases).

Folder Structure

Student_Performance_Prediction/
├── Dataset/                 # Raw and processed data, download scripts
├── Notebook/                # Jupyter notebook for ML pipeline
├── Scripts/                 # Training and utility scripts
├── Model/                   # Serialized models and scalers
├── Streamlit_App/           # Web application and pages
├── Documentation/           # Reports and presentation contents
├── requirements.txt         # Project dependencies
├── README.md                # Project documentation
└── .gitignore

Installation Steps

1. Clone the repository.
2. Create a virtual environment: python -m venv venv
3. Activate the virtual environment:
   • Linux/Mac: source venv/bin/activate
   • Windows: venv\Scripts\activate
4. Install dependencies: pip install -r requirements.txt

Running Instructions

1. Download data: python Dataset/download_data.py
2. Open Notebook: jupyter notebook Notebook/Student_Performance_Prediction.ipynb

Future Scope

• Integration with external LMS (Learning Management Systems).
• Real-time streaming data ingestion.
• Extending predictions to specialized departmental subjects.

student_performance_prediction