期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：3
页码：605
出版社：Journal of Theoretical and Applied
摘要：Medical diagnosis systems play a vital role in medical practice and are used for diagnosis and treatment by several medical practitioners. Diagnosing the risk factors of pre-diabetic (IGT) cases is quite difficult. There is a big challenge to improve the diagnosis system to recognize the risks factors of impaired glucose tolerance regarding to cardiac vascular disease. In this paper, ELM classifier is combined with the hybrid of genetic algorithm and pulse coupled neural network (GENPCNN). Especially, a Single-hidden layer feed forward neural networks are suitable for solving the complex classification problem. The datasets we collected from health care centre having 270 instances of pre-diabetic, Diabetic and non-diabetic data each was having 28 attributes. A combination of genetic algorithm based neural networks to select the features from the dataset. So, it will be reduced to 14 attributes. The best population of the GA will be passed as input for the PCNN. The features extracted from the GENPCNN are passed to ELM classifier SLFNs in which the hidden nodes are chosen randomly and logically determines the output weight. First, dataset is preprocessed in order to remove the noisy data, missing values or irrelevant values and also from ‘curse of dimensionality’ which have to make suitable for training. This algorithm tends to provide good generation performance and extremely fast learning speed. The classification accuracy obtained using this approach is 94%. The obtained results have shown very promising outcomes for the prediction of risk factors of CVD in impaired glucose tolerance and impaired fasting glucose.
