A Parameter Tuned Ensemble Model for Accurate Prediction of Diabetic Progression

The use of ensemble based machine learning technique for medical data to predict disease or health condition of patients is in the spot light for its social impact. The method helps the doctors to prescribe medication and lifestyle changes in early stages to prevent any fatality in patients. Here, hyper parameter tuning based optimization is performed on diabetic progression data, for an XGBoost Ensemble model. The model has given the best score of -0.55 as the negative log loss, for 200 estimators with a learning rate of 0.01. The result obtained is promising and the technique could be extended for other applications as well.