Many works have been done to reduce complexity in terms of time and memory space. The feature selection process is one of the strategies to reduce system complexity and can be defined as a process of selecting the most important feature among feature space. Therefore, the most useful features will be kept, and the less useful features will be eliminated. In the fault classification and diagnosis field, feature selection plays an important role in reducing dimensionality and sometimes might lead to having a high classification rate. In this paper, a comprehensive review is presented about feature selection processing and how it can be done. The primary goal of this research is to examine all of the strategies that have been used to highlight the (selection) selected process, including filter, wrapper, Meta-heuristic algorithm, and embedded. Review of Nature-inspired algorithms that have been used for features selection is more focused such as particle swarm, Grey Wolf, Bat, Genetic, wale, and ant colony algorithm. The overall results confirmed that the feature selection approach is important in reducing the complexity of any model-based machine learning algorithm and may sometimes result in improved performance of the simulated model.