Frequent itemset mining is one of the standard methods for association analysis today. Working on large collections of set-valued transactions, it enumerates all subsets of items that occur at least a given minimal times in transactions. The main obstacle is the exponential number of possible subsets, so that special search strategies have to be employed.

We implemented the three main algorithms for this problem, Apriori, Eclat, and Fp-Growth in C++ using generic programming for optimizing speed. For each base algorithm, many algorithmic features are available for configuration, covering a large number of variants.

Developers: Ferenc Bodon, Balazs Racz, Lars Schmidt-Thieme

Download: fim_env.tar.bz2 (v1.0.3 from March 14, 2006)

Related publications:

• Ferenc Bodon, Lars Schmidt-Thieme (2005):
  The Relation of Closed Itemset Mining, Complete Pruning Strategies and Item Ordering in APRIORI-based FIM algorithms, in Proceedings of the 9th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD) 2005, Porto, Portugal.
• Ferenc Bodon, Balazs Racz, Lars Schmidt-Thieme (2005):
  On Benchmarking Frequent Itemset Mining Algorithms, in Proceedings of the Open Source Data Mining Workshop, 11th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) 2005, Chicago, USA.
• Lars Schmidt-Thieme (2004):
  Algorithmic Features of Eclat, in Proceedings of the Frequent Itemset Mining Implementations Workshop, IEEE International Conference on Data Mining (ICDM), Brighton, UK.

See also: Ferenc Bodons page for our library