minsup

Minimum support, abbreviated minsup, is a user-defined threshold used in frequent itemset mining. In a dataset of transactions, each itemset has a support value representing the proportion (or count) of transactions that contain the itemset. An itemset is considered frequent if its support is at least the minsup value. Support can be expressed as a relative proportion of all transactions or as an absolute count.

Minsup is a central parameter in algorithms such as Apriori and FP-Growth. It serves to prune the search space: only itemsets whose support meets or exceeds minsup are carried forward, and in Apriori, all non-empty subsets of a frequent itemset must themselves be frequent (the Apriori property). This pruning helps manage the combinatorial explosion of potential itemsets in large datasets.

Choosing an appropriate minsup involves trade-offs. A high minsup yields a small set of very common itemsets and fewer association rules, reducing computation but potentially missing interesting patterns. A low minsup increases the number of candidate itemsets and rules, which can capture rarer patterns but also introduces noise and higher computational cost. Threshold selection is often guided by domain knowledge, dataset size, and practical objectives; it can be tuned through exploratory analysis or grid searches.

Practical notes: minsup is not a measure of statistical significance or quality of a rule; it merely filters by frequency. It can be specified as a percentage of the total transactions or as a fixed transaction count. In some implementations, relative support and absolute support are used interchangeably, with the threshold converted accordingly.