Leveraging the Firefly Algorithm for Dimensionality Reduction and Accurate Diabetes Prediction
Keywords:
Firefly Algorithm, Feature Selection, Diabetes Prediction, Machine Learning, Dimensionality Reduction, Metaheuristic OptimizationAbstract
Accurate and early diagnosis of diabetes remains a significant challenge in healthcare, compounded by the high dimensionality of medical datasets. These datasets often contain numerous irrelevant or redundant features, which can negatively impact model performance and increase computational complexity. Feature selection, the process of identifying the most relevant features, is critical in addressing these challenges. This study explores the application of the Firefly Algorithm (FA), a metaheuristic optimization technique inspired by the natural behavior of fireflies, for efficient feature selection in diabetes prediction models. The FA simulates fireflies' attraction behavior to navigate complex search spaces and identifies key features that improve model accuracy. Using a diabetes dataset, the study evaluates the performance of the Firefly Algorithm compared with traditional feature selection methods. Seven classification models were tested, and performance metrics, including accuracy, precision, recall, and F1-score, were used to evaluate the results. The findings demonstrate that the Firefly Algorithm effectively reduces dimensionality while maintaining or improving prediction accuracy. This method shows great potential to enhance diabetes diagnosis by developing more efficient and accurate machine learning models, ultimately leading to improved patient outcomes. The study also highlights the advantages of the Firefly Algorithm over conventional techniques in healthcare applications.
