Risk Prediction of Disease Complications in Type 2 Diabetes Patients Using Soft Computing Techniques

Abstract

Diabetes has become the fourth leading cause of death in developed countries. By the endurance and hasty spread of diabetes, with increased number of ill condition, complications in the disease all over the world, several methodologies have been developed to predict and prevent this chronic disease. An early diagnosis of disease helps patients and medical experts to reduce the problem, risk and cost of medications. This paper presented an efficient system to predict diabetes and further complications with risk level. In this system, methods including genetic algorithm, nearest neighbor, and fuzzy rule-based system have been used in order to provide an accurate prediction system to prepare for presence of diabetes and complications. In this system, 235 individual's data were collected. The best subsets of features generated by the implemented algorithm include the most common risk factors such as age, family history, BMI, weight, smoking habit, alcohol habit and also factors related to the presence of other diabetes complications considered for predication of disease. The proposed system was prejudiced and the results showed to be more suitable by selecting best subset of features selected using variations of genetic algorithm depending on the types of nearest neighbor. The succeeded results produced 95.83% sensitivity, 95.50% accuracy and 86.95% specificity on impenetrable data which support the effectiveness of the system to predict the disease.