CatBoost Approach for Diabetic Disease Classification

Diabetes is a chronic disease in which the pancreas is unable to produce or release enough insulin into the blood, leading to increased blood sugar levels. Diabetes is characterized by increased glucose levels. If it is not properly managed, it damages organs such as the liver, heart, kidneys, and eyes, and it can lead to stroke, hypertension, and cardiovascular diseases. Techniques to monitor blood sugar levels play a vital role in preventing and mitigating diabetes-related consequences. This research aims to classify diabetes disease, which is largely a result of the complexity and heterogeneity of the disease, which emphasizes the need to develop creative classification approaches and advanced diagnostic frameworks. Therefore, the CatBoost classifier is chosen in this research work as it is an effective gradient boosting technique that can handle categorical features directly without preprocessing. It prevents overfitting and performs better on unbalanced target variables. Using this, we have to classify the diabetes disease dataset by splitting the population with different variable ratios. After that, the model achieved a significant accuracy of 98.78% using hyperparameter tuning such as 500 iterations, 0.1 learning rate, and 0.65 SMOTE ratio.