Machine Learning Approach for Diabetes Prediction using Genetic Algorithm based Feature selection

Diabetes, a prevalent and complex medical condition, demands accurate predictive models for early detection and effective management. This paper introduces a novel approach for diabetes prediction by combining genetic algorithm-based feature selection with ML classification. By combining the genetic algorithm's capability to optimize feature selection and the predictive power of ML classifier, this work offers a promising avenue for improving diabetes prediction accuracy. Two datasets from Kaggle were collected. Initially, RF applied on both datasets. Later, datasets ae balanced using oversampling technique "ADASYN". Later, genetic algorithm is employed to optimize feature selection, with the fitness function minimizing the negative accuracy of the model. The selected features are then used to train a final model, and the accuracy is evaluated on the test set. The results showcase the effectiveness of the proposed approach in enhancing diabetes prediction accuracy when compared to base model. Results from both datasets shown accuracy enhancement with GA feature selection. The selected features provide valuable insights into the influential factors contributing to diabetes outcomes.