OPTIMIZED FEATURE SELECTION FOR LUNG CANCER CLASSIFICATION USING NATURE-INSPIRED ALGORITHMS

Abstract

Lung cancer is still one of the most fatal forms of cancer around the world. Early detection is indeed a crucial aspect that can help improve survival rates in cancer patients. Modern imaging and genetic technologies have made it feasible to have access to a wide variety of datasets associated with the detection of lung cancer. But unwanted or redundant features can act as a deteriorating factor that might reduce the performance level of classification methods. Therefore, appropriate feature selection methods are required to curate dimensionality, efficiency, and accuracy in cancer detection methods as well. Meta-heuristic search methods, inspired by nature, have proven to be a powerful aid in overcoming difficult optimization tasks, including feature selection problems in particular. Their stochastic nature makes them more apt to deal with health-related databases efficiently. In this context, this research aims to use two meta-heuristic optimization methods, Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), to find out optimal feature sets from a lung cancer database to judge performance associated with feature sets identified by selecting suitable features that are gauged with K-Nearest Neighbours classification technique individually. Notably, PSO attained a classification accuracy of 97.82%, outperforming GA with 96.74%. These findings underscore the effectiveness of metaheuristic optimization-based feature selection in elevating the diagnostic performance of lung cancer classification systems.