A RDF model-based predictive model on dengue patients' diagnostic PDF reports

Abstract

Resource description framework (RDF) models are increasingly used for storing data in the dengue healthcare domain on account of its explicit knowledge representation capability and ability of keeping provenance. This machine-interpretable graph-structured data model leads to new challenges for the knowledge engineering community to discover unobvious information leading to the necessity of designing RDF-model-based predictive model. The prerequisite is to extract diagnostic information from the patients' diagnostic reports available in portable document format (PDF) format in the hospital and generate RDF model facilitating machine interpretability. Hence, in this work, RDF model is generated as a prerequisite by extracting the clinical information available in the respective PDF reports of 118 dengue patients' using Docparser and RDF123 open source software. A classifier using artificial neural network (ANN) is designed on the top of the RDF model developed providing healthcare domain with an interoperable predictive model. The proposed classifier is evaluated using parameters such as accuracy, specificity, sensitivity (recall), precision, F1-score, etc. along with confidence interval and p-value. It offers a viable solution towards clinical decision making in semantic web-based dengue healthcare domain.