Discovery of interesting association rules using genetic algorithm with adaptive mutation

Association rule mining is the process of discovering useful and interesting rules from large datasets. Traditional association rule mining algorithms depend on a user specified minimum support and confidence values. These constraints introduce two major challenges in real world applications: exponential search space and a dataset dependent minimum support value. Data analyzers must specify suitable dataset dependent minimum support value for mining tasks although they might have no knowledge regarding the dataset and these algorithms generate a huge number of unnecessary rules. To overcome these kinds of problems, recently several researchers framed association rule mining problem as a multi objective problem. In this paper, we propose ARMGAAM, a new evolutionary algorithm, which generates a reduced set of association rules and optimizes several measures that are present in different degrees based on the datasets are used. To accomplish this, our method extends the existing ARMGA model for performing an evolutionary learning, while introducing a reinitialization process along with an adaptive mutation method. Moreover, this approach maximizes conditional probability, lift, net confidence and performance in order to obtain a set of rules which are interesting, useful and easy to comprehend. The effectiveness of the proposed method is validated over a few real world datasets.