A Rule-Based Monitoring System for Accurate Prediction of Diabetes: Monitoring System for Diabetes

1. Introduction

Diabetes s is one of popular chronic disease that can affect the life of people throughout the world. The main reason of diabetes is lack of production of insulin in human body. It can also characterize as metabolic disease and having high blood sugar rate. The blood sugar level varies due to insulin emission. Further, it can also affect the other body parts of human and becomes worse, if cannot treat timely. The reduced level of insulin can also communicate blood glucose in blood platelets. In turn, the chance of several other diseases like heart disease, paralysis, strokes can also increase in diabetic people. In developed countries, diabetes become well-known reason of death and it one of most common non-communicable diseases throughout world (World Health Organization, n.d.). Till 2025, 300 million people will be either diabetic or pre-diabetic in world. A study reveals that fifty million people in India are either pre-diabetic or diabetic (Alice & Balachandran, 2015). Further, this study also highlights that nearly forty-four people in India are not know, they are diabetes affected. Diabetes can occur at any stage in human being. Clinically, diabetes can be categorized in type 1, type 2, and gestational diabetes (Zheng et al., 2007). Type 1 diabetes can be described as juvenile diabetes and it can occur in children frequently, but it also occurs in adults. In Type 1 diabetes, the body is unable to produce insulin in appropriate amount and also destroy the cell which is responsible to produce insulin. Type 2 diabetes can occur any stage of human life, but it mainly occurs in fat, middle-aged, and elderly people. Gestational diabetes occurs in women due to hormonal and other changes. In the literature, numerous algorithms are presented for predicting diabetes. These algorithms include various conventional machine learning methods like decision tree, logistic regression, neural network, etc. (Kavakiotis et al., 2017; Yadav et al., 2012; Kumar & Sahoo, 2013). Some meta-heuristic algorithms and feature selection methods are also reported to determine the risk factors of diabetes disease (Yue et al., 2008; Duygu & Esin, 2011; Sahoo & Kumar, 2014; Gambhir et al., 2016).

It is also noticed that some ensemble methods are also developed for diabetes diagnosis (Kavakiotis et al., 2017; Ozcift & Gulten, 2011; Gambhir et al., 2017; Kumar et al., 2019; Gambhir et al., 2018). So, it is concluded that there is not a single system available in literature that can analyze the personal health record of patients. Hence, there is need of a diagnostic system that can effectively diagnose patients with diabetes. In this work, a diabetes monitoring system is proposed based on personal health record of patients. Further, several machine learning techniques are also used to compare the results of proposed system.