Spam classification has become an essential task due to the rapid increase in unsolicited email communications. This task is challenging because it involves large datasets with high-dimensional feature spaces, where many features are redundant or irrelevant. Traditional feature selection methods often struggle to handle feature redundancy and uncertainty effectively, which can degrade classifier performance. In this paper, we apply rough set theory to spam classification to address these challenges. We apply two rough set-based feature selection methods that aim to identify the most relevant features while reducing dimensionality and preserving classification accuracy. These methods leverage the concept of indiscernibility from rough set theory to filter out irrelevant features and improve the performance of classification models. We compare the performance of these rough set-based feature selection methods with traditional techniques using various classifiers, including Decision Tree, KNeighbors, and others. Our experiments show that the rough set-based feature selection methods outperform traditional methods, achieving higher classification accuracy and significantly reducing the feature set.