Towards Medical Ontology Construction Using Data Mining: An approach for creating a diabetic ontology using clustering

Abstract

Ontologies are abstract representation of domain knowledge that encompasses the structures and relations between concepts. They are stored in a form that prompts sharing, reusing, and querying of the knowledge base. Ontologies use processing and reasoning technology to derive information implied by knowledge. In healthcare, intelligent decision support systems are increasingly employing ontologies for diseases diagnosis, prevention, and treatment. Early diagnosis of diseases such as diabetes helps prevent the progression of severe health problems. Due to the massive amount of diabetes-related data in the medical field, data mining and semantic techniques have been utilized in building automated systems for diabetes prediction and diagnosis. This work aims to implement a methodology for creating a fuzzy ontology for diabetes diagnosis. It proposes a method for constructing a fuzzy ontology based on data mining techniques to reduce the time and effort of ontology construction process. Expectation maximization (EM) clustering algorithm was applied on a diabetic dataset to group concepts and attributes.