Generation of Frequent Itemset with Bit Stream Mask Search and Sparse Bit Mask Search
Abstract
In general, computer systems are often stored as large amounts of data from which a particular record must be retrieved in accordance with some search criterion. As a result, the well-organized storage scheme to smooth the progress of fast searching is a vital issue. Frequent pattern mining was initially proposed by Agrawal et al. for the purpose of market basket analysis in the form of Association Rule Mining (ARM). Researchers have formulated several algorithms for the generation of frequent itemsets. Frequent itemsets are found mainly from the dataset through a number of searching algorithmic schemes. The novel bit search technique is implemented in the existing ARM algorithms. Frequent itemsets are generated with the assistance of apriori based bit search technique is known as Bit Stream Mask Search and eclat based bit search technique is branded as Sparse Bit Mask Search. These two algorithms are executed on six datasets namely T10100K, T40I10100K, Pump, connect-4, mushroom and chess. These six datasets again executed on AprioriTrie and FP-Growth algorithms. All the algorithms are executed in 5% to 25% support level and the results are compared for the purpose of performance evaluation of the proposed scheme. Efficiency is proved through performance analysis.
Keywords Association Rules, Frequent Itemset Mining, Bit Search, Bit Stream Mask Search, Sparse Bit Mask Search.
Full Text