<p>The selection of relevant features is a critical step in many machine learning and data analysis tasks, as it can significantly impact the performance and interpretability of the resulting models. In this work, we introduce a novel feature selection approach that draws inspiration from the unique movement patterns of the narwhal, a fascinating marine mammal, and the power of binary optimization. The proposed method, named BNO-FS, is based on Narwhals optimizer (NO) which is a new meta-heuristic recently developed recently and never been tested on feature selection problem. A binary version is proposed in this study. A new fitness function is proposed composed of two important terms: the classification accuracy rate obtained by three classifiers and the number of selected features. The algorithm aims to identify the optimal subset of features that maximizes the predictive performance of the model while minimizing the number of selected features. The accuracy is measured using SVM, KNN, and decision tree classifiers to account for different learning biases. Averaging across these models helps reduce selection bias and ensures that the selected features are robust and generalizable, rather than overfitted to a single classifier. The effectiveness of the binary Narwhals optimizer approach is demonstrated through a series of experiments on benchmark datasets, where it is compared to other state-of-the-art feature selection techniques. The experimental result shows that BNO-FS achieved a mean classification accuracy of 97.65% on the MIAS dataset, outperforming PSO (97.54%) with a statistically significant improvement.</p>