Applied Machine Learning/2024
Machine Learning for Trading
A quantitative trading study using decision trees, random forests, and Q-learning under conservative evaluation constraints.
PythonNumPyPandasQ-Learning
- in-sample return
- 170%
- out-of-sample return
- 14%
Project note
This study compares decision trees, random forests, and a Q-learner in a quantitative trading workflow. It covers feature construction, policy learning, transaction-aware evaluation, and the gap between in-sample and out-of-sample behavior.
The reported strategy achieved a 170% cumulative in-sample return over two years and 14% out of sample under a conservative setup. The more useful takeaway is methodological: strong training performance is not treated as evidence of generalization, so both periods remain visible in the result.