ATTRIBUTES BASED BEHAVIOR PATTERN ANALYSIS OF ASSOCIATION MINING ALGORITHMSUSING GROUNDWATER DATA

Abstract

One of the most widely used data mining methods is frequent itemset mining. Apriori, Eclat and FP-Growth are the predominantly used algorithms in association mining. A lot of analysis has been going on to establish the relative supremacy of the algorithms when it comes to scalability with respect to increase in the data size. Although there are ample papers that talk about data size already available, a majority of them have never explored the performance of dataset containing unique itemset characteristics.This paperexamines two important data attributes that can have a drastic impact on algorithm performance. Empirical study has been carried out with groundwater dataset and the roles of frequent item density and transaction size with respect to performance were monitored. In addition to gauging performance, the study also helps us understand the dataset in its relation to the algorithm used to study it. The results show that both Eclat and FP-Growth handle impact of transaction size and itemset density increase far better than the Apriori algorithm.