Development of Fast and Scalable Algorithms for Data Mining Technique

Abstract

Data mining is the task of discovering interesting patterns from large amount of newlinehistorical data. Among different data mining techniques, association rules mining became newlineone of the most frequently used technique due to its wide application area which is also newlinefocused in this research work. newlineIn the association rule mining, Frequent Itemset Mining (FIM) is a necessary step newlinefor interesting patterns within databases. The Apriori downward closure property makes newlineit possible to successfully mine sparse datasets. But in the case of dense datasets, the newlinesearch space becomes exponential in the number of items occurring in the database and newlinethe targeted databases tend to be massive, containing millions of transactions. Therefore newlineit may generate a huge number of frequent itemset, especially when minimum support newlinethreshold is set low. Such characteristics affect efficiency of association rule mining and newlinestudied extensively by different research groups. newlineTaking note of these characteristics, an attempt is made and presented here with in newlinethis thesis. In this work, approaches were developed with fast and scalable algorithms for newlinemining frequent closed itemset (FCI) to solve the problem of huge number of frequent newlineitemset. newlineClosed itemsets were chosen as an alternative of frequent itemset to improve the newlinemining efficiency since they are orders of magnitude fewer than frequent itemset, newlineespecially when a dataset containing highly correlated transactions. Furthermore, they newlineconcisely represent exactly the same knowledge as that of frequent itemset. Additional newlineadvantage reflected in association rules generation, where it has been established that newlinethey are more meaningful for analysts, since each of all redundancies is discarded. newlinexii newlineThis research work was planned to use different approaches. In this work set of newlinethree approaches with the improvement one over other are presented. newlineInitially, in the first phase, graph based bottom up approach for frequent closed newlineitemset mining, Concurrent Edge Prevision and Rear Edge Pruning Approach