Specification of Healthcare Expert Systems Using a Multi-mechanism Rule-extraction Pipeline

The application of knowledge extraction methodologies in support of medical informatics promises interesting developments that could potentially improve many aspects of healthcare services. In this paper we outline a multi-stage rule extraction pipeline for rule-based knowledge discovery. The featured methodology would facilitate operationally straightforward extraction of symbolic rules from medical datasets, in particular those with unannotated ordinal or continuous-valued datavectors. The extracted rulesets will be used in the construction or enrichment of rule-based expert systems. Our pipeline incorporates well-established supervised and unsupervised machine learning methods used for data mining. The motivation for our work stems from the individual effectiveness of data mining methods available for datavector clustering, attribute discretisation and rule extraction. The featured knowledge extraction architecture will be tested and analysed using several wellknown medical datasets.