A knowledge discovery pipeline for medical decision support using clustering ensemble and neural network ensemble

It is widely recognized that knowledge discovery and data mining in the health domain are two techniques than scientists and researchers are always looking into areas for improvements and accurateness in prediction. In this paper, we present a multi-tier knowledge acquisition, amalgamation and learning info-structure for the learning of rules that have been generated from medical datasets comprising both annotated and un-annotated attributes. We propose a hybridized approach for rule learning manifested in an enhanced Knowledge Discovery Pipeline which features data cleansing, a novel data clustering ensemble mechanism via boosting, data discretization, rule generation via rough sets, rule filtering and eventually neural network ensemble via bagging. The pipeline, in addition to generating decision rules, would produce a neural knowledge base that can be considered an abstraction of knowledge that is present in the dataset.