Abstract:
Diabetes is a chronic disease caused by elevated blood sugar levels. Its adverse effects in-clude complications such as cardiovascular diseases, kidney failure, and impaired vision,contributing to a substantial health burden to the community. The early predictions of di-abetes play a crucial role in significantly reducing the associated risks and severity. How-ever, the limited availability and confidential nature of data pose substantial challenges toresearchers in predicting diabetes.Additionally, the presence of outliers or missing val-ues in diabetes datasets adversely affects algorithm performance.This research analyzesand investigates the challenge of precise early diabetes prediction by adopting ensemble-based Machine Learning (ML) models.In this study, three different datasets originatingfrom Bangladesh,India,and Germany undergo a thorough investigation using a varietyof ensemble-based and baseline ML models from a cross-country perspective.The out-comesrevealedthatthedatasetfromBangladeshexhibitedsuperiorperformancebyutiliz-ing boosting ML algorithms, including AdaBoost, CatBoost, Gradient Boost, and XGBoost.However, the PIMA dataset from India demonstrated good performance for ML models likeCatBoost, SVM, and Random Forest. The dataset acquired from Germany projected goodperformances for Boosting as well as baseline ML models.The combined dataset is also ex-ploredandtheBoostingalgorithmsperformbettercomparedtootherbaselineMLmodels.The study identified the optimal features for diabetes prediction from the datasets and un-derscored the significance of understanding diabetes variations by exploring the strengthsand weaknesses of ensemble models for three different datasets from three different ge-ographical locations.The study mainly focused on diabetes prediction in the context ofBangladeshi data set over publicly available data and found that the ensemble and cross-countryapproachenhancesthereliabilityofdiabetespredictionwithpotentialimplicationsapplied to diagnose the diabetes of the patients of Bangladeshi people as well as in globalhealthcarestrategiesacrossdiversepopulations.