Optimizing Diabetes Diagnosis Through Pulse Waveform Analysis and Data Mining

The objective of this study is to develop a robust diabetes diagnosis model by employing four distinct algorithms as weak learners: the Random Forest algorithm, SVM algorithm, KNN algorithm, and Decision Tree algorithm. The selection of the optimal classification model involves a meticulous process, and further refinement is conducted through the application of the Stacking classifier, with the Multilayer Perceptron (MLP) classifier serving as the final model. The performance of the optimized model is thoroughly evaluated to identify the most effective diagnostic model. The experiments are conducted using a dataset obtained from Changhua Christian Hospital in Taiwan. Our experimental results show that the performance of the model optimized by the Stacking ensemble learning method is significantly improved. The optimized model achieves an F₁ score of 0.86 and an AUC score of 0.90, indicating the effectiveness of the proposed model.