Pollution-Predictor: A Real-Time Air Quality Forecasting using Deep Learning and Optimization

Overview

This research project focuses on the prediction of environmental pollution levels—specifically air quality, PM2.5 concentration, and temperature-driven pollution dynamics—through the use of historical environmental datasets. To enhance predictive performance, the study employs the Jaya Optimization Algorithm, which is utilized to fine-tune machine learning models. The algorithm improves both accuracy and stability of predictions while maintaining computational efficiency.

Unlike conventional optimization techniques, Jaya is parameter-independent, eliminating the need for algorithm-specific control parameters. This simplicity, coupled with its robustness, makes Jaya particularly well-suited for real-world environmental prediction tasks.

The proposed framework demonstrates potential as a scalable and adaptable solution for pollution forecasting, contributing to applications in environmental monitoring, smart city infrastructure, and evidence-based policymaking.

Features

• 📊 Predicts pollution levels using historical datasets
• ⚡ Optimized with Jaya Optimization Algorithm for enhanced accuracy
• 🌐 Built with TensorFlow.js for real-time, browser-based predictions
• 📈 Scalable to different pollution datasets (air, water, noise, etc.)
• 🔮 Useful for environmental monitoring, smart cities, and decision-making

Repository Structure

📂 ML-Research-Project/
│   ├── requirements.txt
│   ├── package.json      
│   ├── package-lock.json         
│   └── .gitignore
│ 
├── 📁 backend/           
├── 📁 data/              
├── 📁 frontend/      
├── 📁 notebook/      
├── 📁 paper_sections/     
├── 📁 presentation/              
├── 📁 report/           
├── 📁 results/
├── 📁 src/          
└── README.md            

Tech Stack

Languages: Python, JavaScript
Frameworks/Libraries: TensorFlow.js, Scikit-learn
Optimization: Jaya Algorithm
Visualization: Matplotlib, D3.js

---

Installation & Setup

This project has two main components:

1. Python (Machine Learning Research) – for data preprocessing, analysis, and model training.
2. Node.js (Web Application with TensorFlow.js) – for serving the frontend and enabling real-time predictions.

1️⃣ Clone the Repository

git clone https://github.com/rutvikbarbhai/Pollution-Predictor-A-Real-Time-Air-Quality-Forecasting-using-Deep-Learning-and-Optimization.git
cd Pollution-Predictor-A-Real-Time-Air-Quality-Forecasting-using-Deep-Learning-and-Optimization

Python Setup (Machine Learning Research)

1️⃣ Create a Virtual Environment
python -m venv venv

2️⃣ Activate the Environment 
venv\Scripts\activate  

3️⃣ Install Dependencies
pip install -r requirements.txt

4️⃣ Run ML Scripts
python src/roc_plot.py

Node.js Setup

1️⃣ Install Dependencies
npm install

2️⃣ Start the Server
node backend/server.js 

3️⃣  Open in Browser
http://localhost:3000

👨‍🔬 Inventors

• Dr. Tusar Kanti Mishra
  Associate Professor
  Computer Science and Engineering Department
  Manipal Institute of Technology, Bengaluru
  Manipal Academy of Higher Education, Manipal, India
  Email ID: tusar.mishra@manipal.edu

• Rutvik Avinash Barbhai
  Undergraduate Student
  Computer Science and Engineering Department
  Manipal Institute of Technology, Bengaluru
  Manipal Academy of Higher Education, Manipal, India
  Email ID: rutvik.mitblr2022@learner.manipal.edu
  Contact Number: +91 7887367708

• Sheetal Sinha
  Undergraduate Student
  Computer Science and Engineering Department
  Manipal Institute of Technology, Bengaluru
  Manipal Academy of Higher Education, Manipal, India
  Email ID: sheetal.mitblr2022@learner.manipal.edu
  Contact Number: +91 9962008641

• Ankit Sarkar
  Undergraduate Student
  Computer Science and Engineering Department
  Manipal Institute of Technology, Bengaluru
  Manipal Academy of Higher Education, Manipal, India
  Email ID: ankit3.mitblr2022@learner.manipal.edu
  Contact Number: +91 8700879300

Future Work

• Extend to multi-pollution datasets (air, water, noise, soil)
• Deploy as a cloud-based monitoring tool
• Integrate with IoT sensors for live environmental data streaming