A New Frequent Pattern Mining Algorithm with Weighted Multiple Minimum Supports

Abstract

Association rules mining is one of the momentous areas in data mining. Frequent patterns mining plays an important role in association rules mining. The effects of traditional frequent patterns mining with same minimum support are highly affected by the value of minimum support. But, for many real datasets, it’s hard to choose the value of minimum support. Too small values of minimum support may cause rules explosion, and too large values may cause rare item dilemma. In this paper we propose an improved approach to extract frequent patterns, which are more interesting to users. Because of the different characteristics of each item, we assign a multiple minimum support and weight based on item support and users’ interests for each item. In order to define the minimum supports of itemsets, we suggest a novel method, which exploits the minimum constraint and maximum constraint to deal with the rare item dilemma and rules explosion problem. The combination of minimum constraint and maximum constraint is based on the weight of the itemset. In this way, we extend the support confidence framework. Experimental results show that the proposed approach is more efficient than other comparing methods.

KeyWords:

• Data mining
• Association rules
• Frequent pattern
• Interest
• Multiple minimum supports

Acknowledgements

This work was supported by the Chongqing 151 Major Science and Technology Projects (cstc2013jcsf-jfzhX0002), Chongqing, China.