查看文章

Diabetes Mellitus (DM) is considered as a heretical metabolic disorder and widely spread long standing slow poison which poses a great threat to human health. Faster and accurate prediction of diabetes is a dire need and Machine Learning (ML) can play a pivotal role in terms of enhancing medical health technology and develop an e-healthcare system. In this regard, ten ML algorithms have been studied comprehensively and they are implemented by Jupyter Notebook. Hence, the ML models are trained with the dataset of Kaggle machine learning data repository of Frankfurt hospital, Germany. Effective data processing method is proposed using 5-fold cross validation method to achieve stable accuracy. However, hyper-parameter tuning technique is employed with a view to achieving better performance from the ML models. After rigorous simulation, Gaussian Process (GP) emerged as the best performing algorithm which is proposed as the most efficient classifier with an accuracy of 98.25%. However, Random Forest (RF) and Artificial Neural Network (ANN) displayed accuracy of 97.25% and 96.5% respectively which are quite satisfactory. Hence, the performances of the ML models are assessed with different metrics like Accuracy, Sensitivity, Precision, F1-score, Specificity and ROC_AUC and thus, a comparative analysis among all the ML models are portrayed graphically. Efficient prediction of Diabetes by ML algorithms can significantly contribute in decreasing the annual mortality rate specially in developing countries like Bangladesh. Therefore, this study can meaningfully assist the healthcare professionals in the process of proper and …