Enhanced Biased Weights AdaBoost Algorithm for Diabetes Detection on Imbalanced Dataset

The paper proposes a new approach to improve multi-fault diabetes diagnosis using ensemble learning - biased weights with AdaBoost algorithm on unbalanced datasets. The suggested technique makes utilization three distinct using machine learning algorithms like the Support Vector Machine,, Logistic Regression and Decision Trees. These models are combined using the AdaBoost algorithm with or without bias weights. The performance of the approach is evaluated using an imbalanced diabetes dataset of 2000 samples, and the results are compared to various machine learning models, namely Naive Bayes, Among the methods employed are kNearest Neighbours and Support Vector Machine with radial basis function, Artificial Neural Network, and popular ensemble learning techniques such as XGBoost, Light Gradient Boosting Machine, and Random Forest. Diabetes diagnosis accuracy is significantly improved by the proposed method, according to the study’s findings, particularly when using decision tree models and unbalanced datasets. The most elevated precision pace of 99% is accomplished utilizing AdaBoost with decision tree as weak learner, alongside unfairness loads of 0.7 and 0.4. The findings suggest that the proposed technique could be effective for identifying diabetes in clinical settings. It could also be used to create a decision support system for doctors to help them diagnose and treat diabetes.