Manickam, Selvakumar
(2013)
An Ontology-Driven Methodology To Derive Cases From Structured And Unstructured Sources.
PhD thesis, Universiti Sains Malaysia.
Abstract
The problem-solving capability of a Case-Based Reasoning (CBR) system largely depends on the richness of its knowledge stored in the form of cases, i.e. the CaseBase (CB). Populating and subsequently maintaining a critical mass of cases in a CB is a tedious manual activity demanding vast human and operational resources. The need for human involvement in populating a CB can be drastically reduced as case-like knowledge already exists in the form of databases and documents and harnessed and transformed into cases that can be operationalized. Nevertheless, the transformation process poses many hurdles due to the disparate structure and the heterogeneous coding standards used. The featured work aims to address knowledge creation from heterogeneous sources and structures. To meet this end, this thesis presents a Multi-Source Case Acquisition and Transformation Info-Structure (MUSCATI). MUSCATI has been implemented as a multi-layer architecture using state-of-the-practice tools and can be perceived as a functional extension to traditional CBR-systems. In principle, MUSCATI can be applied in any domain but in this thesis healthcare was chosen. Thus, Electronic Medical Records (EMRs) were used as the source to generate the knowledge. The results from the experiments showed that the volume and diversity of cases improves the reasoning outcome of the CBR engine. The experiments showed that knowledge found in medical records (regardless of structure) can be leveraged and standardized to enhance the (medical) knowledge of traditional medical CBR systems. Subsequently, the Google search engine proved to be very critical in “fixing” and enriching the domain ontology on-the-fly.
Actions (login required)
 |
View Item |
