Feature selection using game Shapley improved grey wolf optimizer for optimizing cancer classification
摘要
Cancer has become a leading global health challenge, significantly impacting mortality rates worldwide. Early diagnosis of cancer can substantially reduce patient mortality and improve treatment success. In recent years, machine learning models have played a vital role in cancer classification, enhancing the precision and efficiency of diagnostic strategies. However, existing feature selection methods are often wrapper-based and computationally intensive, especially with high-dimensional biomedical data. Addressing this gap, we propose a two-stage hybrid algorithm for efficient and accurate feature selection, combining kernel Shapley Value (kSV) and Improved Grey Wolf Optimization (IGWO). In the first stage, kSV is used, which leverages calculation and combination of feature contributions by considering interactions among features. In the second stage, IGWO optimizes the selection process by modeling it as an optimization problem. The proposed kSV-IGWO algorithm was tested on eight benchmark cancer datasets using a support vector machines (SVM) classifier. Experimental results demonstrate that the kSV-IGWO algorithm outperforms conventional methods, achieving high accuracy on benchmark datasets