A novel approach to design a domain specific deep learning ontology

Abstract

Chronic diseases are considered as a major hurdle for human life across the newlineglobe. Machine Learning and Deep Learning techniques have been extensively used newlinein medical field to predict and diagnose chronic diseases. Early disease prediction newlineand diagnose these diseases will reduce the maximized severity of having further newlineseverity of the disease and hence associated mortality. newlineThe main objective of this research is to propose a method that involves newlineontology development for a specific and significant domain. The title of this work is newlinebroader in nature and its scope is limited for medical field. Hence the ontologies are newlinedeveloped for medical datasets by providing a valid relationship between the newlineconcepts and attributes. It improves the classification, accuracy and at the same time newlinereduces the computational time. In order to achieve the desired objectives, this newlineresearch first proposes an efficient ontology development with semantic web rule newlinelanguages. The key aspect is the selection of attributes based on the uniqueness of newlinethe dataset and also validation of the considered dataset through rules formulate newlineusing semantic web rule language. The research undertaken work is suitable for newlinevarious types of data that contains in the dataset that show the remarkable results newlinewith different chronic disease dataset. newlineSecondly a novel ontology based disease prediction system by employing newlinemachine learning algorithms is proposed for the prediction of chronic diseases. The newlinegeneralization performance of machine leaning algorithms depends on the dynamic newlineselection of appropriate features to perform efficiently. These algorithms perform newlineusing ontology based approach and 10 fold cross validation to obtain the best results newlinefor the dynamic features that are based on specificity, sensitivity, accuracy and newlineMatthew correlation coefficient. newline newline