Hybrid Feature Selection and MLP Classifier for Diabetes Prediction

Diabetes mellitus (DM) poses a significant global health challenge, necessitating early detection and precise classification to prevent complications and reduce healthcare burdens. This study analyzes a hybrid machine learning model combining Least Absolute Shrinkage and Selection Operator (LASSO) feature selection with a Multilayer Perceptron (MLP) classifier to ameliorate the prediction accuracy of DM. Applied to a publicly available dataset, the hybrid model underwent rigorous preprocessing and hyperparameter optimization, achieving high values for accuracy (98%) and mean F1 scores (0.98) comparable to standalone MLP classifiers. The hybrid approach demonstrated significant advantages in feature reduction and therefore could enhance model interpretability and potential computational efficiency. The results underscore the potential of hybrid models to enhance DM diagnostic tools, facilitating early intervention and improved patient outcomes.