Data Mining Approach for the Early Risk Assessment of Gestational Diabetes Mellitus

Introduction

With increasingly easier access to clinical databases and healthcare medical centers, diagnostic systems are now used in a myriad of medical domains. Large volumes of medical data accumulate in healthcare clinical centers. Having access to these data centers, it is possible to extract and discover tacit knowledge behind the data through analysis of real events and results. Nowadays, in the modern industrial world with changing lifestyle alimental manner and slake mobility the incidence of complex diseases like gestational diabetes is rapidly increasing. Recently, high rates of obesity and diabetes as well as low mobility and increasing marriage age, have resulted in increased occurrence of diabetes. Gestational diabetes mellitus is a kind of diabetes that appears for the first time during pregnancy (Ovesen, Jensen, Damm, Rasmussen, & Kesmodel, 2015). Its prevalence is reported to be 1-3% in the United States, 10.9% in Asian countries, 5.2% in Europe (Coustan, 2013) and ranges from 4-17% in all pregnancies (San, Ling, & Nguyen, 2012). GDM statistics are increasing on a daily basis, due to lifestyle changes in industrial urban areas as well as reduction of mobility and environmental factors. Having access to large amounts of data and database available in medical centers, it is possible to apply data mining methods. In the proposed research, after analysis, the primary data and output data of patients discovered that among the input and output data there were patients that they affect to GDM in future, but they have no initial warning symptoms or primary risk factors. Some data were unknown areas for symptoms and reasons of affection. Therefore, these exceptional cases may generate through encryption aspect of knowledge that are imperceptible for humans. However, human knowledge continues to grow and expand. If GDM is not treated or diagnosed in a timely manner, it may cause serious side effects for both the mother and the child. The diagnostic systems of the various complicate diseases like GDM are rapidly increased in recent years. Certain approaches such as expert systems, fuzzy logic, Artificial Neural Networks (ANNs), and decision-making support systems can help to improve the diagnosis and prediction of the disease. Different features of each approach can be used in diverse situations. For detection of these hide aspects of science and recognize the relationship between data simultaneously ANNs help to have a better and higher accuracy in prediction of complex disease like GDM. Complex diseases are often accompanied by unpredictable effects and are not free from false presumptions. Gestational diabetes mellitus is often diagnosed in middle or late period of pregnancy, when it is too late for prediction and the mother should be under medical treatment. In the proposed approach, the risk of GDM notices to the pregnant mother in the initial stages of pregnancy. However, it is possible by paying attention to this warning and taking a suitable food diet and necessary tendency the prevention of affection to GDM. This paper is designed to answer the question of “What is the best approach for the timely and accurate prediction of GDM?” Many data mining techniques have been developed that are applicable for many domains. For this purpose, the two most referenced, popular and basic methods of classification, the techniques of Artificial Neural Network (ANN) and Decision Tree (DT), are used to reduce the amount of error and make more accurate predictions: