ASSOTSIATIV QOIDALARNI IZLASH USULLARI VA ALGORITMLARI

Abstract

This article analyzes the problem of mining association rules from large transactional databases, including its formal model and computational stages, based on two fundamental scientific sources. The first source discusses support and confidence constraints, frequent itemset discovery, evaluation, pruning, and memory management methods [1]. The second source examines the partitioning of the search space into a lattice and equivalence classes, the vertical TID-list representation, the Eclat, MaxEclat, Clique, and MaxClique algorithms, as well as their experimental comparison with Apriori and Partition [2]. The analysis demonstrates that algorithm selection depends not only on execution time but also on data density, minimum support, the size of the longest frequent itemset, the number of database scans, and main-memory constraints. Vertical data representation and hybrid search can provide significant advantages for dense datasets, whereas a simple level-wise search may be practically sufficient for sparse databases.