A small Python project that uses a contextual bandit to pick the best POP (edge location) per request context to reduce latency.
- Treats each POP as an “arm” and selects one per request.
- Learns from feedback (RTT/latency + error penalties) to balance exploration vs exploitation.
- Supports simple policies like LinUCB and Thompson Sampling.
Python, NumPy, pandas (scikit-learn for baselines), Matplotlib (plots)
git clone https://github.com/rithwik-01/ML-Driven-Routing-Decisions.git
cd ml-driven-routing-decisions
pip install -r requirements.txt
python -m src.simulate --policy linucb