Early Detection of Diabetes Using Multi-Strategy Machine Learning Models: A Smart Healthcare Approach for Enhanced Diagnostic Accuracy

Abstract

Undiagnosed diabetes is very dangerous, and a growing public health issue around the world. We need to be able to identify early if we want to halt the course of the disease and mitigate risks to critical organs. This project will build a multi-faceted ML model that will increase the accuracy and efficiency of diabetes diagnosis. They use a dataset that has health indicators such as age, BMI, insulin and glucose to train a number of machine learning algorithms to classify the risk of diabetes. DenseNet 96.8% was the best algorithm of all algorithms and AlexNet 94.8%. Other good performers included Weighted KNN and Fine Gaussian SVM with 83.4% and 83.5% accuracy respectively. These models were evaluated further on performance metrics such as F1 score, accuracy and recall. For example, because of its ability to detect positive cases, DenseNet recall reached 0.97. This proposed approach combines ensembles and sophisticated feature selection for diagnosing purposes. The model’s clinical fit is guaranteed by statistical calculations used to confirm its fit in the real world. Finally, by offering a robust machine learning-based early diabetes detection solution, this study makes advances in medical diagnostics and aids clinicians to treat patients more quickly and effectively.