Comparative anatomization of data mining and fuzzy logic techniques used in diabetes prognosis

Abstract

Diabetes is an ailment in which glucose level increase in at high rates in blood due to body’s inability to metabolize it. This happens when body does not produce sufficient amount of insulin or it does not respond to it properly. Critical and long-term health issues arise if diabetes is not handled or properly treated which includes: heart problems, disorders of the lungs, skin and liver complications, nerve damage, etc. With increasing number of diabetic patients, its early detection becomes essential. In this paper, our major focus areas are data mining and fuzzy logic techniques used in diabetes diagnosis. Data mining is used for locating patterns in huge datasets using a composition of different methods of machine learning, database manipulations and statistics. Data mining offers a lot of methods to inspect large data considering the expected outcome to find the hidden knowledge. Fuzzy logic is similar to human reasoning system and hence it can handle the uncertainties found in the data of medical diagnosis. These systems are called expert systems. The fuzzy expert systems (FES) analyze the knowledge from the available data which might be vague and suggests linguistic concept with huge approximation as its core to medical texts. In this paper, the methodology section delivers the pipeline of various tasks such as selecting the dataset, preprocessing the data by applying numerous methods such as standardization, normalization etc. After that, feature extraction technique is implemented on the dataset for improving the accuracy and finally dataset worked on data mining and fuzzy logic various classification algorithms. While analyzing different data mining methods, the accuracy computed through random forest classifiers as high as 99.7% and in case of numerous fuzzy logic approaches, high precision and low complexity was found to contribute a fairly high accuracy of 96%.