Analytic Algorithm for Predicting Diabetes Based on GSDRC-Stacking-Anchor Model

“How to prevent diabetes” has become a global issue that is addressed as the incidence of diabetes keeps rising quickly and the number of deaths from complications keeps rising. The majority of researchers right now concentrate on model performance optimization, ignoring the study of decision transparency and interpretability, which limits the field’s ability to be widely applied to highly risk choice scenarios like diabetes prediction. Due to the great internal complexity of the models, the accuracy of highly interpretable models frequently falls short of expectations in terms of accuracy. We propose the GSDRC-Stacking-Anchor predictive parsing model to overcome the aforementioned shortcomings. First, we suggest a data processing strategy based on the DRC Loss function to address the problems of aberrant feature dimension and unbalanced sample distribution in the original data. The GSDRC parameter optimization algorithm, which is based on theDRC Loss function and the grid search method, is proposed in order to address the problems that the traditional grid search algorithm is not applicable to data samples with unbalanced distribution and the tuning parameters time is too long for the model. Last but not least, the GSDRC-Stacking-Anchor interpretable prediction model is proposed, and the Stacking integrated learning method and interpretable Anchor algorithm are introduced on the basis of a single prediction model after parameter optimization. This method can significantly reduce the workload of physicians, help doctors make preventive decisions, and improve patients’ self testing and prevention abilities, as well as provide a direction for some future medical research.