Performance Analysis and Prediction of Diabetes using Various Machine Learning Algorithms

Nowadays, diabetes has become very common. In every house, there should be a person who has diabetes. It is a menacing condition, which means dangerous, and this disease is known as a chronic illness. This happens due to a high level of glucose or sugar in the human body, and then it becomes very indispensable to maintain a level of sugar throughout our lifetime, so it needs to be predicted earlier in the starting days. They are majorly in type 1 and type 2, and gestation is also there, which occurs in pregnancy. Diabetes may cause eye diseases like increasing eye pressure and glaucoma. There may be a chance of heart disease and a slow recovery rate of the pancreas. In this article, we are going to propose and analyze the machine learning algorithms for diabetes prediction by using a dataset of different features (Glucose, BMI, Age, etc.). Outcome of this paper is accuracy, precision, recall and f1-score for various machine learning algorithms. Consideration of implementation on all features, without glucose and without pregnancy for diabetic and non-diabetic. For all features logistic regression has highest accuracy i.e 74.45%, for without glucose feature KNN gives highest accuracy level i.e 68.83% and without pregnancy feature, KNN accuracy with 76.19%.