Using Machine Learning for the Prediction of Diabetes with Emphasis on Blood Content.

Diabetes is a condition associated with high levels of blood glucose and insulin deficiency that is becoming more prevalent among children and adolescents. It is the ninth leading cause of mortality globally at over 1 million deaths per year. Misdiagnosis of type 1 diabetes with type 2 are frequent and causes diabetic ketoacidosis, a critical complication. This paper seeks for the most accurate model between frequently used machine learning algorithms - Multilayer Perceptron, Support Vector Machine and Random Forest in predicting diabetes among patients with focus on blood content analysis, specifically using the Laboratory of Medical City Hospital (LMCH) Diabetes Dataset, which was retrieved from Mendeley. Using hyperparameter optimization and ANOVA F- Value feature scaling, results show that Random Forest produced the highest accuracy at 1.00, followed by Multilayer Perceptron at 0.987 and Support Vector Machine at 0.96. [ABSTRACT FROM AUTHOR]