EduPulse

EduPulse is a student performance analytics project built with React, Flask, Spark, and a lightweight ML pipeline. It shows subject-wise performance, attendance, and at-risk students across CSE, DSAI, and ECE.

The current version focuses on:

• subject-wise total marks
• attendance analysis
• at-risk identification
• branch-level summaries
• branch-wise risk comparison charts

Features

• Subject dropdown for BDA, DL, DSP, and DBMS
• Per-student subject totals on a normalized 100-point scale
• At-risk detection based on marks and attendance
• Overview cards for total students, at-risk count, marks, and attendance
• Branch-wise performance summary
• Search, branch filter, sorting, and pagination for student records
• Search in the at-risk view by student name or ID
• Risk-type filters for Marks < 33% and Attendance < 75%
• Subject-aware analytics charts built with Chart.js

File Storage (HDFS)

All datasets in this project are stored in the Hadoop Distributed File System (HDFS), which is configured and run locally on the system.

Instead of relying on local file storage, the project uses HDFS to simulate a distributed data environment, enabling scalable data processing using Apache Spark.

HDFS Directory Structure

/students_data/
├── students/
│   ├── cse_students.csv
│   ├── dsai_students.csv
│   └── ece_students.csv
│
├── attendance/
│   ├── cse_attendance.csv
│   ├── dsai_attendance.csv
│   └── ece_attendance.csv
│
├── marks/
│   ├── cse_BDA_marks.csv
│   ├── cse_DL_marks.csv
│   ├── cse_DSP_marks.csv
│   ├── cse_DBMS_marks.csv
│   ├── dsai_BDA_marks.csv
│   ├── ...
│   └── ece_DBMS_marks.csv
│
└── processed_data/
    └── final_dataset.csv

## Marks Normalization

Each subject is normalized to `100` marks using this formula:

total_subject_marks =
  quiz1_marks
  + quiz2_marks
  + assignment_marks
  + (mid_sem_marks × 0.4)
  + (end_sem_marks × 0.4)

This means:

• quiz1 remains as-is
• quiz2 remains as-is
• assignment remains as-is
• mid sem is scaled from 50 to 20
• end sem is scaled from 100 to 40

So the final subject total is out of 100.

At-Risk Criteria

A student is marked as at-risk if either condition is true:

• subject total marks are below 33
• attendance is below 75%

In code terms:

selected_marks < 33 OR attendance_pct < 0.75

Because the dashboard is subject-aware, the same student may be safe in one subject and at-risk in another.

Tech Stack

• Frontend: React + Vite + Chart.js
• Backend API: Flask + Pandas
• Data processing: PySpark
• ML: Logistic Regression with Spark MLlib

Project Structure

BDA-Project-4th-Sem/
├── backend/
│   ├── flask/
│   │   └── app.py
│   ├── ml/
│   │   └── train_model.py
│   ├── raw_data/
│   │   ├── *_students.csv
│   │   ├── *_attendance.csv
│   │   └── *_marks.csv
│   └── spark/
│       └── pipeline.py
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   ├── constants/
│   │   └── hooks/
│   ├── .env
│   └── package.json
└── README.md

Backend Overview

1. Spark Processing

backend/spark/pipeline.py reads raw student, attendance, and marks CSV files, computes:

• attendance percentage
• normalized subject totals
• average marks across subjects
• binary at-risk label for the ML dataset

The processed dataset is written to:

hdfs://localhost:9000/students_data/processed_data/final_dataset.csv

2. ML Pipeline

backend/ml/train_model.py trains a Logistic Regression model using:

• avg_marks
• attendance_pct

It outputs prediction data including:

• prediction
• risk_score

3. Flask API

backend/flask/app.py serves the dashboard data. It combines the processed predictions CSV with raw subject-mark files so the frontend can query subject-specific totals.

Available endpoints:

• GET /health
• GET /data?subject=BDA
• GET /at-risk?subject=BDA
• GET /summary?subject=BDA
• GET /branch/<branch>

The default subject is BDA if no subject is provided.

Frontend Overview

The frontend is a React dashboard with these tabs:

• Overview
• All Students
• At-Risk
• Analytics

Main UI behavior:

• a global subject dropdown controls subject-specific marks
• all views update when the subject changes
• marks bars and charts use the normalized 100-point scale
• the Analytics tab shows Safe vs At-Risk by Branch
• the Analytics tab can be filtered by Marks < 33% or Attendance < 75%
• the At-Risk tab can be filtered by risk reason and searched by student name or ID

Environment Configuration

Backend

backend/.env

PROCESSED_CSV=data/predictions.csv
PORT=5000
HDFS_NAMENODE='http://localhost:9000'
SPARK_MASTER='local[*]'

Running the Project

1. Start the Flask backend

From the project root:

cd backend/flask
python3 app.py

The API runs by default on http://localhost:5000.

2. Start the frontend

From the project root:

cd frontend
npm install
npm run dev

The Vite app will run on its local dev port and call the Flask API using VITE_API_URL.

Team

• Aalekh Raghuvanshi
• Bhavya Khare
• Devam Sharma
• Hemant Kumar
• Saksham Kushwah