A risk prediction model using RF-WFS and XGBoost achieved better accuracy and classification performance for type 2 diabetes diagnosis compared to other standard classification algorithms.
Does an ensemble learning model using RF-WFS and XGBoost improve prediction accuracy for type 2 diabetes compared to other classification algorithms?
An ensemble learning model combining random forest-based weighted feature selection and XGBoost improves the accuracy of type 2 diabetes risk prediction.
Type 2 diabetes mellitus is a severe chronic disease threatening human health and has a high incidence worldwide. People need to use effective prediction model to diagnose and prevent diabetes in time. At present, data mining technology has become an increasingly important technology with classification capability in the field of medical diagnosis. This paper proposes a risk prediction model for type 2 diabetes based on ensemble learning method. In the proposed model, the weighted feature selection algorithm based on random forest (RF-WFS) is used for optimal feature selection, and extreme gradient boosting (XGBoost) classifier. The effectiveness of the method was validated by comparing the various performance metrics and the results of different contrast experiments. Additionally, we get a better prediction accuracy using the method than using the other classification algorithms (C4.5, Naive Bayes, AdaBoost, Random Forest). The validation results at UCI Pima Indian diabetes dataset shows that the model has better accuracy and classification performance than other research results mentioned in the literature. As a result, it has been proven that the model would be effective for the diagnosis of diabetes at the initial stage.
Xu et al. (Sat,) conducted a other in Type 2 diabetes mellitus. Risk prediction model based on RF-WFS and XGBoost vs. Other classification algorithms (C4.5, Naive Bayes, AdaBoost, Random Forest) was evaluated on Prediction accuracy and classification performance. A risk prediction model using RF-WFS and XGBoost achieved better accuracy and classification performance for type 2 diabetes diagnosis compared to other standard classification algorithms.