Email spams pose a significant risk to the cybersecurity, and privacy of users. Several studies have utilized various approaches, but not all of them indicated the correct usage of the feature selection technique, along with the class balancing techniques. This paper examines a machine learning method that utilizes an artificial neural network (ANN) including preliminary data feature selection, design, training, and evaluation to classify emails as spam or non-spam. Initially, feature selection is done with Weka’s correlation-based feature selection (CFS) evaluator and Best First search (BFS) method. The results show that the model is effective in spam detection on the UCI Spambase dataset with 93% accuracy, and an area under curve with 97%. This approach addresses class imbalance through an effective class balancing technique, ensuring a more robust model that can accurately distinguish between spam and legitimate emails, and the importance of the feature selection method in enhancing the model performance.