Multi-Layer Neural Network (Mlnn)-Based Classification Of Diabetes

Diabetes Mellitus (DM) is found to be common among people nowadays. Factors like age, obesity, absence of exercise, genetic factors, life style, improper diet, increased blood pressure, etc. cause diabetes. People with diabetes are highly prone to diseases in the heart and kidney, stroke, problem in the eye, damage in the nerve etc., as glucose levels are not accurately controlled. Necessary information for diagnosing diabetes is to be gathered by conducting diverse tests and suitable treatment must be offered depending on diagnosis. The existing schemes do not offer improved classification as well as prediction accuracy. In this paper, a Deep Learning (DL)-based diabetes prediction model is proposed for offering improved classification based on external factors which cause diabetes in addition to other factors including Glucose, Age, BMI, Insulin, etc. The proposed Long Short-Term Memory (LSTM)-based no-prop Multi-Layer Neural Network (LSTM-MLNN) offers improved classification accuracy in contrast to existing methods of classification of DM. Experimental analysis is performed based on sensitivity, specificity and accuracy.