Binary Flying Squirrel Optimizer for Feature Selection

• Luiz Fernando Merli de Oliveira Sementille UNESP
• Douglas Rodrigues UNESP
• André Nunes de Souuza UNESP
• João Paulo Papa UNESP

Resumo

Bio-inspired optimization algorithms aim to address the most diverse problems without the need for derivatives, and they are independent of the shape of the search space. The Flying Squirrel Optimizer belongs to the family of bio-inspired algorithms and simulates the movement of flying squirrels from tree to tree in search of food. This paper proposes a binary version of the flying squirrel optimizer for feature selection problems. To elucidate the performance of the proposed algorithm, we employed six other well-known bio-inspired algorithms for comparison purposes in sixteen benchmark datasets widely known in the literature. Furthermore, we employ the binary flying squirrel optimizer in selecting gas concentrations to identify faults in power transformers. The results expressed that Binary Flying Squirrell Optimizer can either find compact feature sets or improve classification effectiveness, corroborating its robustness.