← All projects
Courseworkcoursework
ML for Trading — CS 7646
A market simulation framework with Q-learning strategy learners.
The problem
CS 7646 wires reinforcement learning and supervised learners into a market simulator. The deliverable is a full simulation framework — not a notebook — with reproducible Sharpe, cumulative return, and drawdown comparisons against buy-and-hold.
Who this is for
Prospective OMSCS students, people considering applying ML to trading and wanting to see the academic version first.
Architecture
- Market sim
- Order execution, fills, holdings, cash over historical prices.
- Q-learning strategy learner
- Discretized state space; reward shaped from return.
- Random Forest strategy learner
- Supervised baseline; same features, different fit.
- Evaluator
- Sharpe, cumulative return, max drawdown vs buy-and-hold.
Request / data flow
- 01Build feature set from price history.
- 02Train Q-learner / RF on training period.
- 03Walk forward over test period through the sim.
- 04Compare to buy-and-hold baseline.
Stack
PythonQ-LearningRandom ForestSharpe RatioRL