Aim: The exhaled biomarkers like volatile organic compounds (VOC’s) are highly potential for the précised diagnosis of diseases. The vital intention of this research is to predict the diabetes mellitus using VOC analysis. Methods: This research work presents an Internet of Things (IoT) prototype for earliest diagnosis of diabetes using breath analysis. The exhaled breath contains different volatile organic compounds (VOCs) generated during metabolic processes in both healthy and pathological conditions. The analysis of breath metabolites provides non-invasive and cost effective diagnosis of human diseases than blood and urine tests. For the exhaled breath analysis, the samples from 348 diabetes patients and 152 healthy volunteers were identified. Totally 500 breath samples were collected over the developed hardware prototype in non-invasive manner. These samples were mined using existing and the proposed classification model for the diagnosis and classification of diabetes mellitus. The existing algorithms Naive Bayes, Logistic Regression, Multilayer Perceptron and AdaBoost are used to classify the developed dataset. Findings and Discussions: The results of the classifiers are compared based on accuracy measures. Amongst these algorithms the Naive Bayes and Logistic Regression algorithm provides maximum classification results. By the combination of these algorithms, a proposed algorithm Bagged Decision Tree is developed. The results obtained by proposed algorithm are compared with the results of existing algorithms. Conclusion: The results found that the proposed algorithm is better than the classification of existing algorithms.
Volume 11 | 06-Special Issue
Pages: 1377-1382