The node provides different algorithms to searches for frequent items in a list of item sets. The integrated algorithms are:
- Apriori (Agrawal et al. 1993)
- FPgrowth (frequent pattern growth, Han et al 2000)
- RElim (recursive elimination)
- SaM (Split and Merge)
- JIM (Jaccard Item Set Mining)
Apriori: The apriori algorithm (Agrawal et al. 1993) carries out a breadth first search on the subset lattice and determines the support of item sets by subset tests. This is a pretty fast implementation that uses a prefix tree to organize the counters for the item sets. The census data set may be used to test the program.
FPgrowth: The FPgrowth algorithm (frequent pattern growth, Han et al 2000) represents the transaction database as a prefix tree which is enhanced with pointers that organize the nodes into lists referring to the same item. The search is carried out by projecting the prefix tree, working recursively on the result, and pruning the original tree. Since version 1.2 this implementation also contains the alpha-pruning of the FP-Bonsai techniques.
RElim: The RElim algorithm (recursive elimination) is inspired by the FP-growth algorithm, but does its work without prefix trees or any other complicated data structures. The main strength of this algorithm is not its speed (although it is not slow, but even outperforms apriori and eclat on some data sets), but the simplicity of its structure. Basically all the work is done in one recursive function of fairly few lines of code.
SaM: The split and merge algorithm (Split and Merge) combines a depth-first traversal of the subset lattice with a horizontal transaction representation. The main strength of this algorithm is not its speed (although it is not slow, but even outperforms apriori and Eclat on some data sets), but the simplicity of its structure. Basically all the work is done in one recursive function of about fairly few lines of code. In addition, it only uses a simple array as the only data structure.
JIM: Finds Jaccard item sets with an extension of the Eclat algorithm. In analogy to frequent item set mining, where one tries to find item sets the support of which exceeds a user-specified threshold (minimum support) in a database of transactions, a Jaccard item set is an item set for which the (generalized) Jaccard index of its item covers exceeds a user-specified threshold. This measure yields a much better assessment of the association strength of the items than simple support. Since the (generalized) Jaccard index is, like the support, also anti-monotone, the same basic approach can be used for the search, provided it is extended to compute the denominator of the Jaccard index.
Dice: Finds Dice item sets with an extension of the Eclat algorithm. In analogy to frequent item set mining, where one tries to find item sets the support of which exceeds a user-specified threshold (minimum support) in a database of transactions, a Dice item set is an item set for which the Dice index of its item covers exceeds a user-specified threshold.
Tanimoto: Finds Tanimoto item sets with an extension of the Eclat algorithm. In analogy to frequent item set mining, where one tries to find item sets the support of which exceeds a user-specified threshold (minimum support) in a database of transactions, a Tanimoto item set is an item set for which the Tanimoto index of its item covers exceeds a user-specified threshold.