Spam email remains a major challenge in the digital world, requiring intelligent and efficient detection methods. This paper introduces a hybrid model that combines the parameter-free Jaya optimization algorithm with the LightGBM boosting method to enhance spam detection performance. Jaya is used to automatically optimize LightGBM’s hyperparameters, improving learning ability without manual tuning. The proposed Jaya-LightGBM model was evaluated on the Enron spam dataset using five-fold cross-validation, achieving 98.07% accuracy, 99.36% recall, 95.64% precision, and 97.41% F1-score, outperforming the baseline LightGBM and several recent approaches. Additionally, the model is integrated with LIME to provide transparency and interpretability. Overall, the findings highlight the effectiveness of combining Jaya optimization with LightGBM for building robust, scalable, and efficient spam detection systems.