Comparing euclidean distance and nearest neighbor algorithm in an expert system for diagnosis of diabetes mellitus

Abstract

Increasing sugar levels in the body which exceeds the normal limit is a metabolic disease commonly called diabetes mellitus. Long-term diabetes mellitus is one of the causes of other diseases such as liver, heart and other body organs. Early diagnosis of diabetes mellitus in a person is very important to know earlier. Early diagnosis is made to prevent other diseases to reduce the occurrence of complications in the body. The use of existing cases can be compared to new cases to diagnose whether the patient has diabetes. One method that can be used is a case-based expert system which is a reasoning system that uses old knowledge to be compared with new knowledge to overcome new problems. This case-based expert system provides a solution based on the similarity of new cases to existing cases. Some methods that can be used to do the similarity process are Euclidean distance and Nearest neighbor. Old cases taken are cases that have the highest similarity value. Result of the similarity value of a case is considered unsuccessful if it is diagnosed or the target case is <80, then the new case will be revised by the expert. The test results show that the system is able to recognize diabetes mellitus using the nearest neighbor similarity method, and Euclidean distance similarity with the calculation of accuracy using the euclidean distance similarity method is 93.33% and the nearest neighbor similarity method of 86.67%, So that the euclidean distance method is more effective because it has a higher accuracy value than the nearest neighbor method.