Frequent Itemset Mining Using LP-Growth Algorithm Based on Multiple Minimum Support Threshold Value (Multiple Item Support Frequent Pattern Growth)

摘要

Frequent itemset mining is an important role in association rule mining. An itemset is called as regular if it fulfills the condition of minimum support threshold values. While using single minimum support for the entire transaction database, the classic model of frequent itemset mining leads to the issue of "infrequent item problem." For higher minimum support, frequent itemset with infrequent items are missed. For the lower minimum support combinative burst can occur, discovering huge number of frequent itemsets. To solve the infrequent item problem, "multiple minimum supports framework" is used to find frequent itemsets. In this idea, minimum item support vale (MIS) is given to each items. The idea of minimum support for an itemset is determined as the minimal MIS value of all its items. Attempts have been made to propose multiple item support based on linear prefix tree known as MIS-LP-Growth to discover frequent itemsets. This concept permits the user to indicate higher minsup for an itemset which contains frequent items only and lower minsup for an itemset which contains rare items only. From the analysis, it is showed that the proposed MIS-LP-Growth is outstanding than the existing LP-Growth algorithm in the criteria of execution time and memory.