Association Rule Mining with Apriori and FP-Growth

Summary

This article introduces association rule mining as an unsupervised method for finding item combinations that recur in transaction data, then explains how the ideas may be applied to trading. It defines support as the frequency of an itemset in the data and confidence as the frequency of a rule relative to its antecedent. The anti-monotonicity property allows infrequent itemsets to be pruned because their supersets cannot occur more often.

A worked example applies Apriori to ten transactions, iteratively generating and filtering candidate sets before deriving rules that meet minimum support and confidence thresholds. The article also describes FP-Growth, which organizes transactions in a compact tree and mines frequent patterns without generating the same breadth of candidates. The treatment is mainly theoretical: the transaction example illustrates the calculations, but the text provides no trading dataset, market test, or evidence that mined associations predict profitable trades. It notes that practical implementation and evaluation are deferred to a later article.

Key ideas

• Association rule mining searches for recurring co-occurrences in transaction data.
• Support measures how often an itemset occurs, while confidence measures how often a consequent follows an antecedent.
• Apriori prunes candidate sets using the rule that a superset cannot be more frequent than its subsets.
• The worked example shows how to build frequent itemsets iteratively and derive rules that pass thresholds.
• FP-Growth uses a tree representation to mine patterns without generating candidates in the Apriori manner.
• The article presents theoretical methods but does not establish their trading profitability.

Tags

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