Combining Unsupervised Learning and Reinforcement Learning for Trading

Summary

This profile outlines a trader’s progression from computer engineering and robotics into systematic trading. It describes an interest in using unsupervised learning to uncover structure in market data, then reinforcement learning to train a neural network through repeated reward-based simulations. The trader also recounts using indicators such as MACD, Stochastics, RSI, and order-book patterns to form directional hypotheses, and seeking a workflow for collecting and processing price, fundamental, and alternative data, generating signals, backtesting, and managing positions.

The material is an autobiographical account of interests and learning, not a documented strategy study. It provides no model specification, validation method, or performance evidence for the proposed machine-learning approach. Its discussion of automation emphasizes executing predefined rules while away from connectivity; the examples do not establish that automation or the proposed learning methods produce an edge.

Key ideas

• The author proposes using unsupervised learning to identify structure in market data.
• Reinforcement learning is described as a way to iteratively train a trading model against a reward function.
• Technical indicators and order-book patterns helped motivate systematic signal generation.
• A practical algorithmic workflow includes data collection, signal creation, backtesting, and risk management.
• The profile presents hypotheses and personal experience rather than validated performance results.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.