PDO-FA: A Novel Hybrid Approach for Feature Selection From Cancer Datasets

Identifying important features from the high-dimensional dataset is a crucial and challenging task for properly diagnosing a patient. In this case, feature selection is a process adopted by researchers to select optimal features by removing redundant ones. This is known as a combinatorial optimization problem. This article presents a novel hybrid feature selection approach called E-PDOFA. It is a hybrid approach that combines Praire-Dog Optimization (PDO) and the Firefly algorithm (FA). An S-shaped transfer function converts the data as a binary version of PDO. The candidate solutions obtained by the PDO are fed to FA to select the most optimal feature subset. This is the main motto for proposing a novel hybrid algorithm; it increases the exploitation of the search space and enhances the overall performance of the machine learning model. The proposed algorithm has experimented with different cancer datasets and various existing approaches. The performance of the proposed model is quite adequate for identifying the most optimal feature with better accuracy of 99.87% for SRBCT dataset.