Feature selection (FS) plays a crucial role in classification tasks. In this chapter, we consider feature selection as a multi-objective optimization problem and propose a Problem-Specific Non-dominated Sorting Genetic Algorithm (PS-NSGA) (Zhou et al. in Inf Sci 547:841–859 (2021), [1]). In PS-NSGA, we have designed an accuracy-prioritized dominance operator, which increases the survival chances of individuals with higher classification accuracy in the population. Concurrently, a mutation retry operator and a combination operator were also designed to enable our algorithm to converge faster and more effectively. Finally, we developed a solution selection strategy to identify the most suitable feature subset. We conducted a series of experiments on 10 real-world high-dimensional datasets. The results demonstrate that PS-NSGA demonstrates outstanding performance in reducing the number of features while maintaining competitive classification accuracy, compared to some of the state-of-the-art evolutionary and traditional FS algorithms.