<p>Epileptic seizure detection is crucial for clinical diagnosis and ongoing patient monitoring, especially for those with chronic epilepsy. EEG-based seizure detection leverages the brain’s electrical signals to capture temporal and spatial variations that characterize seizure events. However, seizure patterns are often non-linear, patient-specific, and temporally dynamic, making their detection complex and error-prone. Current methods that use machine learning and deep learning models are constrained by challenges such as irrelevant feature inclusion, overfitting, poor generalization, and ineffective management of high-dimensional EEG data. To address these challenges, a novel model is proposed in this research work by incorporating Coati Optimization Algorithm-Based Feature Selection with Convolutional Sparse Autoencoder for accurate and efficient epileptic seizure detection. The proposed model aims to reduce feature redundancy through biologically inspired optimization and improve learning through sparse, high-level feature representations. The methodology was evaluated on the benchmark UCI Epileptic Seizure Recognition Dataset, and experimental analysis ensures the model robust performance with an accuracy of 96.5%, sensitivity of 95.2%, specificity of 97.3%, precision of 96.0%, F1-score of 95.6%, and an AUC of 97.8%. These results of proposed model outperform traditional classifiers such as CNN, SVM with PCA, Random Forest, k-NN, and Deep Autoencoder across all metrics. The integration of optimized feature selection with sparse convolutional encoding contributed significantly to the model’s ability to generalize and minimize false detections, offering a promising tool for automated and reliable seizure detection in clinical and real-time environments.</p>