Classification of Diabetes using Machine Learning

• Published in: 2021 International Conference on Computational Performance Evaluation (ComPE) pp. 185 - 189
• Main Authors: Islam, Nair Ul, Khanam, Ruqaiya
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2021
• Subjects: Classification; Classification algorithms; Diabetes; Healthcare; Kernel; Machine Learning; Machine learning algorithms; Performance evaluation; Reliability; Support vector machines
• DOI: 10.1109/ComPE53109.2021.9751955

Diabetes is an autoimmune disorder which can affect people of any age group or gender. Some people inherit it genetically and others develop it at some point in their lives. In any case it can bring hardship and misery to the patient, which can be both mental and physical in nature. One good way of combating diabetes could be to make its diagnosis quicker, cheaper, and affordable. By making full use of modern day computing capabilities that thing is very much possible. To make diagnosis of diabetes we employ a variety of Machine Learning algorithms like Decision Trees (DT), Naive Bayes (NB), Support Vector Machine (SVM), Logistic Regression (LR). The idea behind making use of many algorithms is for the performance evaluation of the commonly applicable Machine Learning algorithms, and to gauge which algorithm could best fit our need. An accuracy of 78.5% was achieved for SVM with the polynomial kernel and 77.9% with the Linear Kernel. Also, Gaussian Naive Bayes achieved a classification accuracy of 79.87%.