🛜 5G Network Monitor - ML Analytics Platform

ML-powered real-time network monitoring and analytics for telecommunications engineers

A full-stack ML application showcasing real-time 5G network performance monitoring with machine learning-powered anomaly detection, coverage classification, and predictive analytics.

🎯 Project Overview

This project demonstrates end-to-end ML engineering skills for the telecommunications industry:

• Real Network Data: Uses Ookla Open Data + synthetic 5G KPIs (RSRP, RSRQ, SINR, CQI)
• Production ML Models: Anomaly detection, coverage classification, time-series forecasting
• Dual Platform Deployment: React dashboard (GitHub Pages) + Gradio ML app (HuggingFace Spaces)
• Interactive Visualizations: Real-time charts, geographic heatmaps, performance analytics

✨ Key Features

📊 React Dashboard

• Real-time Metrics: Live throughput, latency, signal strength, network quality
• Time-Series Analysis: Interactive Chart.js visualizations of network performance
• Geographic Coverage: City-by-city network performance mapping
• Hourly Analytics: Performance trends by time of day
• Dark Mode: Professional UI with Material-UI

🤖 ML Analytics (Gradio)

• Anomaly Detection: Isolation Forest model detecting network issues (5% contamination threshold)
• Coverage Classifier: Random Forest predicting signal quality (99.95% accuracy)
• KPI Prediction: LSTM forecasting throughput/latency (MAE: 2.3 Mbps)
• Interactive Demos: Try ML models with custom network parameters

🏗️ Architecture

├── frontend/              # React + Vite dashboard
│   ├── src/
│   │   ├── components/   # Chart components, metrics cards
│   │   └── App.jsx       # Main dashboard
│   └── public/           # Data files (JSON)
│
├── gradio-app/           # ML analytics app
│   ├── app.py           # Gradio interface
│   └── requirements.txt
│
├── ml/                   # Machine learning models
│   ├── anomaly_detector.py    # Isolation Forest
│   ├── coverage_classifier.py # Random Forest
│   ├── kpi_predictor.py       # LSTM predictor
│   └── models/               # Trained models (.pkl)
│
├── data/
│   ├── raw/              # Original datasets
│   └── processed/        # Cleaned data
│
└── scripts/              # Data processing utilities

🚀 Quick Start

Prerequisites

• Docker Option: Docker Desktop
• Manual Option: Python 3.9+, Node.js 18+, npm or yarn

Option A: Docker (Recommended - Easiest Setup)

Just run one command to start everything:

./start.sh

This will:

• Build and start the React dashboard at http://localhost:5173
• Build and start the Gradio ML app at http://localhost:7860
• Show backend logs in the terminal

To stop:

./stop.sh

Note: ML models and data must be generated first (see step 2 in Manual Setup below if starting fresh).

---

Option B: Manual Setup

1. Clone Repository

git clone https://github.com/YOUR-USERNAME/telecom-network-monitor.git
cd telecom-network-monitor

2. Generate Data & Train Models

# Install Python dependencies
pip install -r requirements.txt

# Generate synthetic network data
python scripts/create_synthetic_data.py

# Train ML models
python ml/anomaly_detector.py
python ml/coverage_classifier.py
python ml/kpi_predictor.py

# Prepare data for frontend
python scripts/prepare_frontend_data.py

3. Run React Dashboard

cd frontend
npm install
npm run dev

Visit http://localhost:5173 to see the dashboard!

4. Run Gradio ML App

cd gradio-app
pip install -r requirements.txt
python app.py

Visit http://localhost:7860 for interactive ML analytics!

📊 Machine Learning Models

Anomaly Detection

• Algorithm: Isolation Forest
• Features: RSRP, RSRQ, SINR, CQI, throughput, latency, packet loss
• Performance: Detects 5% of samples as anomalies
• Use Case: Real-time network health monitoring

Coverage Classifier

• Algorithm: Random Forest (100 trees)
• Classes: Excellent, Good, Fair, Poor
• Accuracy: 99.95%
• F1 Score: 0.9995
• Top Features: RSRQ (41%), SINR (25%), RSRP (19%)

KPI Predictor

• Algorithm: LSTM (2 layers, 64/32 units)
• Target: Throughput & latency forecasting
• MAE: ~2.3 Mbps
• Use Case: Predictive capacity planning

📈 Datasets

Ookla Open Data

• Source: teamookla/ookla-open-data
• Metrics: Download/upload speed, latency
• Coverage: Global network performance tiles
• Usage: Geographic analysis, baseline performance

5G Synthetic Dataset

• Based on: Irish 5G Dataset schema
• Samples: 50,000 time-series points over 30 days
• Metrics: RSRP, RSRQ, SINR, CQI, throughput, latency, packet loss
• Scenarios: Excellent, Good, Fair, Poor, Anomaly

🎨 Technology Stack

Frontend

• Framework: React 18 + Vite
• UI: Material-UI (MUI)
• Charts: Chart.js, react-chartjs-2
• Styling: Emotion CSS-in-JS

ML/Backend

• ML: scikit-learn, TensorFlow/Keras
• Data: pandas, numpy
• Viz: Plotly, matplotlib
• API: Gradio 4.16

Deployment

• Dashboard: GitHub Pages (static hosting)
• ML App: HuggingFace Spaces (Gradio SDK)
• CI/CD: GitHub Actions

📦 Deployment

GitHub Pages (React Dashboard)

cd frontend
npm run build
# Deploy dist/ folder to GitHub Pages

HuggingFace Spaces (Gradio App)

1. Create new Space at huggingface.co/new-space
2. Select Gradio SDK
3. Upload:
   • gradio-app/app.py
   • gradio-app/requirements.txt
   • ml/ folder
   • data/raw/ sample data
4. Space will auto-deploy!

🎓 Skills Demonstrated

For ML/AI Roles in Telecom:

• ✅ Real 5G network data analysis
• ✅ Time-series forecasting (LSTM)
• ✅ Anomaly detection (unsupervised learning)
• ✅ Multi-class classification
• ✅ Feature engineering for telecom KPIs
• ✅ Model deployment (HuggingFace, web apps)
• ✅ Full-stack ML engineering
• ✅ Data visualization & dashboards
• ✅ Domain knowledge (3GPP standards)

📝 License

MIT License - see LICENSE file

👤 Author

Bita Rahmat Zadeh

• Portfolio: My Portfolio
• LinkedIn: My LinkedIn
• Github: @bitarah

🙏 Acknowledgments

• Ookla for open network performance data
• IEEE DataPort for 5G datasets
• HuggingFace for Spaces hosting
• Gradio for ML app framework

---

⭐ Star this repo if you find it useful for your telecom ML projects!