IBAVOA: An Improved African Vulture Optimization Algorithm for Feature Selection

In the domain of machine learning, redundant and irrelevant features will significantly impede a model's learning ability. However, identifying an effective subset of features from a massive pool of feasible solutions remains a challenge. To address this issue, this paper introduces IBAVOA, an optimization algorithm designed for feature selection. Built upon the excellent performance of the African Vulture Optimization Algorithm, this algorithm enhances its capabilities through initialization with tent map, an adaptive quasi-oppositional mutation strategy, and the S-shaped function. Comparative experiments conducted on six datasets from UCI repository against five popular optimization algorithms demonstrate that this algorithm can effectively reduce feature dimensions while simultaneously achieving a better classification accuracy.