Railways recently implemented Preventive Track Maintenance (PTMs) to enhance the condition of railway tracks, but they failed to concentrate on real-world demands or emergencies on railway tracks. Furthermore, few estimation methods have been tested on real-world- world datasets and incorporated with only binary classification labels actually results in poor classification results. These strategies are challenged in the case of multi-class or multi-variant data. In addition, multi-class categorizations using unsupervised learning that encounters issues, find patterns automatically from the unlabelled data. The dataset involved in this research work is comprised of many features thus making it difficult to select appropriate techniques based on Machine Learning (ML) in industrial systems. Relatively small data sets may result in time loss and unfeasible maintenance scheduling. Thus, FS strategies help to reduce computation time, improve prediction accuracy, and understand the data better. As the traditional strategies, exhibit weaker performance of ML methods due to their dependency on features. The Ensemble Feature Selection (EFS) procedure is envisioned as a machine learning task that requires a suitable input dataset. Evolutionary Computing (EC) strategies have already been used for feature selection and categorization because of their overall optimization capabilities.