Performance Analysis of Classical and Evolutionary Algorithms for Mining Association Rules

Association rule mining techniques based on support and confidence, generates a large number of rules - many of which are useless. As a result, post analysis is needed. For such an optimization problem evolutionary algorithms can be used as they use objective functions that measures the interestingness of rules better. We apply classical and evolutionary algorithms on different types of datasets. Though none of the algorithms is superior than the others - some of the rules obtained by evolutionary algorithms, could not be obtained by classical algorithms. Time is also a great issue that get affected significantly by support value. To our knowledge, there is no work to compare classical (Apriori and FP-growth) and evolutionary (EARMGA and MOEA) algorithms focusing on their performances in different phases of execution. In this work, detailed information is specified according to some of the major components with various aspects. The whole process is treated from data mining perspective and discussed the issues responsible for affecting the performance of the algorithms.