Using FP-Growth Association Rules to Discover Trading Feature Patterns

Summary

The article explains how FP-Growth can mine associations among binary features in historical trading data. Unlike Apriori, which repeatedly scans the database to evaluate candidate patterns, FP-Growth builds a tree representation and performs subsequent mining from that in-memory structure. The author proposes including target outcomes alongside indicator or instrument features, then searching for rules associated with those outcomes.

The implementation uses a multi-branch tree whose nodes store feature IDs, support values, and parent links. The mining logic is adapted to focus on target features and to account for paths where a target appears before later nodes. An Expert Advisor was tested on real data under settings used in earlier tests. The author reports that it did not identify every fractal correctly, while describing the results as interesting. The excerpt does not provide quantitative performance measures or enough detail to assess predictive value, robustness, or profitability; association discovery alone does not establish a tradable edge.

Key ideas

• FP-Growth reduces repeated database scans by storing transactions in a tree for subsequent mining.
• Association rules can be sought among binary features drawn from indicators, instruments, or time intervals.
• The proposed trading application adds target outcome features to the training data and mines rules connected to them.
• The implementation focuses rule construction on selected target features and accounts for later nodes along relevant tree paths.
• The reported Expert Advisor test is qualitative and does not establish profitability or generalization.

Tags

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