Enhanced Detection of Diabetes Mellitus using Prominent Machine Learning Algorithms

Diabetes Mellitus, the scientific name for diabetes, is a chronic metabolic disease that has various health effects on individuals all over the world. Diabetes is like a silent killer that is imposing a threat to human life. The health issues caused by diabetes are heart diseases, strokes, kidney diseases, blindness, etc. The reasons for this are due to sedentary lifestyles, irregular routines, weight gain, lack of proper nutrition, increased cholesterol levels, aging, genetic mutations, and other associated conditions. This increase in diabetes cases facilitates the need for an accurate diabetes diagnosis. For this, some supervised machine learning (ML)algorithms are employed in the diagnosis of diabetes from vast kinds of data with enhanced performance. In this research, ML algorithms like Linear Regression (LR), Decision Tree (DT), Support Vector Machine (SVM), k-nearest neighbors (KNN), Random Forest (RF), Gradient Boosting, and XGBoost, using a widely accessible dataset. To address outliers and values missing, the dataset is preprocessed using methods like imputation, feature scaling, and normalization. Additionally, feature engineering is used to include new categorical variables for BMI, insulin, and glucose levels, improving the model's prediction ability. we used metrics like F1-score, recall, AUC-ROC, accuracy, and precision to determine our model’s performance through hyperparameter transformation and cross-validation. To provide better results, our system demonstrates how important data for the improving of the model results. This study demonstrates the advantageous benefits of machine learning in healthcare for diabetes diagnosis. These findings pave way for future developers in decision-making and predictive analytics for the treatment of chronic diseases.