Machine learning models, at their core, are software systems that demand computational resources, making them a pertinent concern for software engineers. The energy consumed by these models during training and inference phases can have far-reaching consequences, from the environmental impact to operational costs and even the user experience. Consequently, understanding how different software components, such as fea- ture selection methods, impact energy efficiency becomes essential for software engineers tasked with building sustainable and cost- effective AI-driven solutions. By addressing four key research questions (RQ1-RQ4), we aim to provide software engineers with actionable insights into making informed decisions about feature selection methods to strike the right balance between energy efficiency and model accuracy.