The primary motto of Feature Selection is maximizing relevance whilst achieving minimum redundancy. It generally proceeds by determining a feature subset with relevant features. There may be many features that contributes towards decision making process in determining a maintenance action. At times certain features may not contribute towards decision making and they are termed as noisy features. Therefore, the process of Feature Selection helps to remove those geometrical features that does not contribute towards determining the maintenance actions necessary for a track segment. Elimination of non-relevant and highly noisy features by preserving maximal relevance with minimal redundancy to the intended target variable. Reduction of overall computational complexity, training and testing time of the predictor or classifier, resulting in the development of highly cost-effective models. Enhancing the algorithms’ learning performance, mitigating the impact of overfitting, thus ultimately developing better learning models.