An efficient slime mould algorithm combined with K-nearest neighbor for medical classification tasks

Growing science and medical technologies have produced a massive amount of knowledge on different scales of biological systems. By processing various amounts of medical data, these technologies will increase the quality of disease detection and enhance the usability of health information systems. The integration of machine learning in computer-based diagnostic systems facilitates the early detection of diseases, enabling more productive treatments and prolonged survival rates. The slime mould algorithm (SMA) may have drawbacks, such as being trapped in minimal local regions and having an unbalanced exploitation and exploration phase. To overcome these limitations, this paper proposes ISMA, an improved version of the slime mould algorithm (SMA) hybridized with the opposition-based learning (OBL) strategy based on the k-nearest neighbor (kNN) classifier for the classification approach. Opposition-based learning improves global exploratory ability while avoiding premature convergence. The experimental results revealed the superiority of the proposed ISMA-kNN in various classification evaluation metrics, including accuracy, sensitivity, specificity, precision, F-score, G-mean, computational time, and feature selection (FS) size compared with the tunicate swarm algorithm (TSA), the marine predators algorithm (MPA), the chimp optimization algorithm (ChOA), the moth-flame optimization (MFO) algorithm, the whale optimization algorithm (WOA), the sine cosine algorithm (SCA), and the original SMA algorithm. Performance tests were run on the same maximum number of function evaluations (FEs) on nine UCI benchmark disease data sets with different feature sizes.