سیستم هوشمند ترکیبی جدید با محاسبه ارزش از دست رفته برای تشخیص دیابت A novel hybrid intelligent system with missing value imputation for diabetes diagnosis

1. Introduction Recently, diabetes becomes the widespread and major disease in the world. Several researchers have focused on this disease to reduce the growth of this problem. People become diabetics when their body either does not produce or appropriately use insulin. Indeed, the insulin as a hormone plays crucial role in this disease. Sugar and other food are converted into required energy by insulin. Genetics and environmental factors are important reasons of this disease. In 1980, 108 million have diabetes while 422 million people have diabetes worldwide [1–3]. Two types of diabetes are defined- juvenile diabetes and adult-onset diabetes [2]. The obesity is a main cause of adult diabetes that can be postponed or controlled with appropriate diet and exercise, but complete cure of diabetes isn’t possible. It is hard to diagnose the diabetes due to the presence of several factors. Doctors commonly judge by evaluating the current test results of a patient or by comparing the patient with other patients that had similar symptoms and test results. Consequently, recognition of diabetes is very complicated issue for doctors [2]. Hence, scientists and researchers have tried to present an intelligent diagnostic system for diagnosing diabetes. The aim of this paper is to introduce and investigate the method for making a novel robust intelligent diagnosis system based on training data with missing values and lower dimension of the clinical attributes. Among various fuzzy modeling techniques with continuous output such as Adaptive Network Based Fuzzy Inference System (ANFIS), this paper proposed a hybrid classifier based on Logistic regression and ANFIS named LANFIS with binary output to help the physician on diagnosis of diabetes disease. The diagnosis performance of the LANFIS intelligent system is estimated using sensitivity, specificity, accuracy and confusion matrix.