A Low-Complexity, Low-Memory Frequent Itemset Mining Algorithm for Transactions With Sorted Items
7 Pages Posted: 20 Jan 2019 Last revised: 16 Aug 2019
Date Written: June 2, 2019
In this paper, we present an algorithm for finding frequent itemsets in transaction databases in which items are consistently ordered within the list of transactions. The algorithm is designed to reduce redundant reads from the database, and to minimize storage requirements. The algorithm reads through the items in each transaction one at a time and strategically distributes transaction ID's among sublists, that reduce the size of subsequent searches. The data structures used to store these sublists are simple tables whose sizes can be pre-specified, thus making it unnecessary to allocate memory dynamically (which costs additional computer cycles). The tables contain very little redundant information, so that memory is used very efficiently; and table entries are automatically overwritten when no longer needed.
Keywords: association rules, itemsets, mining, frequent, database, complexity, memory
JEL Classification: C80, C88
Suggested Citation: Suggested Citation