NetRouteAI — Optimising Network Paths with AI

NetRouteAI simulates a real-life 5G network topology and, using ML models trained on simulated datasets, predicts node health and link latency. It includes a real-time D3.js dashboard and computes paths in real time using four strategies: Lowest Risk, Fastest (Latency), Least Hops, and Best (Hybrid).

---

✨ Features

• Real-time 5G-like topology simulation (gNB, CORE_DC, EDGE_DC, FIREWALL, etc.).
• Live metrics per node/link (CPU, memory, latency, packet loss) with smoothing & spikes.
• ML training & prediction:
  • Latency via XGBoost Regressor.
  • Node health / alarm via classifiers (e.g., RandomForest/XGBoost).
• D3.js topology visualization with health & latency overlays (predicted + real).
• Real-time Socket.IO stream for computed analysis.
• AI pathfinder strategies: lowest risk, fastest, least hops, best (hybrid).
• Flask endpoints to start/stop analysis and compute paths.

---

🧱 Architecture (high level)

Node.js simulator  --->  JSON/CSV files  --->  Python trainers (XGBoost / SKLearn)
       |                       |                     |
       |                       v                     v
       |               graph_live.json     trained models (*.joblib)
       |                       |                     |
       +----> Flask (app.py) <----------------------/
                 |   \__ launches predictors & generator
                 | 
                 +--> D3.js dashboard (templates/index.html)
                 |
            SocketIO (src/aiAnalysis.py on :5050)

Ports

• Flask app (app.py): http://127.0.0.1:5000/
• AI Analysis SocketIO (src/aiAnalysis.py): http://127.0.0.1:5050/

---

🧰 Tech Stack

Backend: Python (Flask, Flask-SocketIO, NetworkX), scikit-learn, XGBoost, pandas, numpy
Frontend: HTML/CSS/JS, D3.js
Generator: Node.js (json2csv), CSV exports for training
Data: JSON under static/graph-data/ (e.g., graph.json, graph_live.json, predicted files)

---

🎬 Demo Video & GIFs

Watch NetRouteAI in action:

GIF Previews

Intro Screen:

Real-time 5G topology visualization:

AI Analysis dashboard with predicted latency and node health:

Path finder with different criteria (auto-refresh paths):

📦 Requirements

• Python 3.9+
• Node.js 16+ (for the simulator & json2csv)
• pip & virtualenv recommended

Python packages (from requirements.txt):

blinker==1.9.0
click==8.2.1
colorama==0.4.6
Flask==3.1.1
flask-cors==6.0.1
itsdangerous==2.2.0
Jinja2==3.1.6
joblib==1.5.1
MarkupSafe==3.0.2
networkx==3.5
numpy==2.3.1
pandas==2.3.0
python-dateutil==2.9.0.post0
pytz==2025.2
scikit-learn==1.7.0
scipy==1.16.0
six==1.17.0
threadpoolctl==3.6.0
tzdata==2025.2
Werkzeug==3.1.3
xgboost==3.0.2

Latency training snippet uses:

import pandas as pd
import glob
from xgboost import XGBRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error, r2_score
import joblib

---

🚀 Setup & Run (Local)

1️⃣ Clone & Python setup

git clone https://github.com/<your-username>/NetRouteAI.git
cd NetRouteAI

python -m venv venv
# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

pip install -r requirements.txt

2️⃣ Install Node.js (if not already installed)

Download from https://nodejs.org/ and install.

Check version:

node -v

If your repo has a package.json, run npm install.
If not, install the needed package explicitly: npm i json2csv.

3️⃣ Start the backend & frontend

python app.py

This:

• Starts Flask backend on http://127.0.0.1:5000/
• Launches mock data generator (src/patterned_mock_graph_generator.js)
• Runs ML predictors (src/predict_latency.py & src/predict_alarm_status.py)

4️⃣ Start AI Analysis dashboard

In a new terminal (with your virtualenv active):

python src/aiAnalysis.py

This starts the SocketIO server on port 5050 for real-time AI analysis updates.

5️⃣ Open the UI

• Main dashboard: http://127.0.0.1:5000
• SocketIO service: http://127.0.0.1:5050 (programmatic)

---

🔁 Key Runtime Files

• Source (base) topology: static/graph-data/graph.json (editable)
• Live graph (real): static/graph-data/graph_live.json (written by simulator)
• Predicted latency: static/graph-data/graph_live_predicted.json (written by predictor)
• Predicted alarms: static/graph-data/graph_live_alarm_predicted.json (written by classifier)
• CSV exports: csv-data/*.csv (rotated & used by training scripts)

---

📡 Core Endpoints (Flask in app.py)

• POST /ai-analysis/start → spawn src/aiAnalysis.py
• POST /ai-analysis/stop → terminate running analysis process
• POST /predict-path
  Request JSON:

  { "source": "nodeA", "target": "nodeB", "strategy": "best" }

  Where strategy ∈ ["lowest_risk", "fastest", "least_hops", "best"] (default: "best").

---

🧪 Tips

• Ensure static/graph-data/ exists and contains a valid graph.json export.
• If files are mid-write, the analysis server uses safe JSON loading with stability checks/retries.
• If you change ports, update your frontend fetch/socket targets accordingly.

---

📁 Project Structure (simplified)

NetRouteAI/
├─ app.py
├─ src/
│  ├─ aiAnalysis.py                # SocketIO server (port 5050)
│  ├─ patterned_mock_graph_generator.js
│  ├─ generate_startup_data.py      # One file that invokes mock js file and prediction backend
│  ├─ predict_latency.py
│  ├─ predict_alarm_status.py
│  ├─ train_model.py
│  ├─ train_alarm_classifier.py
├─ templates/
│  └─ index.html
├─ static/
│  ├─ js files			# front-end js 
│  ├─ images 			# Icons
│  └─ graph-data/
│     ├─ graph.json
│     ├─ graph_live.json
│     ├─ graph_live_predicted.json
│     └─ graph_live_alarm_predicted.json
├─ csv-data/                       # created at runtime
├─ requirements.txt
└─ README.md

---

License

⚠️ This project is for personal, educational, and portfolio/demo use only. See LICENSE for details.

---

👤 Credits

Created by Mohit Gautam.