FAST ASSOCIATION RULE MINING USING DATA STRUCTURE

Abstract

In the field of data mining, frequent pattern mining has become an important task and has many utilities in various areas. One of the methods to find these frequent patterns for large datasets is using distributed data mining where the database is stored in a distributed way at multiple sites. FDM (Fast Distributed Mining) is one of the many algorithms available for distributed data mining which uses the Apriori algorithm to find the local candidate itemsets at each of the site for a number of passes. In recent years, many new structures like, Node-list, N-list and Nodesets have been introduced to find the frequent itemset in a more efficient way which uses both the tree structure of FP-Tree and the intersection property of vertical datasets. So, a new algorithm named as FDM-FIN (FDM using FIN) is proposed by using the Nodeset structure in place of array and hash tree structures in the FDM algorithm to generate and store the candidate itemsets by applying the FIN algorithm to generate the candidate itemsets locally at each site instead of using the apriori algorithm. The performance of this algorithm is then compared to the FIN algorithm and FDM using FP Growth where FP Growth is used to generate the candidate sets.