Dynamic PSO-Based Associative Classifier for Medical Datasets
Veenu Mangat, Renu Vig · IETE Technical Review · 2014
Association rule mining is a data mining technique for finding actionable and novel patterns from databases. Rules are simple to understand and can be easily used for classification. For such a system to be widely adopted, it should give output that is correct and comprehensible. With the explosion of information, databases have become humongous containing a large number of records and attributes. Traditional methods of classification need to be modified to give output that is satisfactory to end users. This paper discusses a rule mining classifier based on a Dynamic Particle Swarm Optimizer. Due to its seeding procedure, concept of regions and their regrouping, it disallows premature convergence and provides a better value in every dimension. Quality evaluation is done both for individual rules and for entire rule sets. Experiments were conducted to evaluate the performance of the proposed algorithm in comparison with other state-of-the-art associative classifiers. Results demonstrate competitive performance of the proposed method.