Epilepsy is a neurological disease of the brain that causes repeated seizures. Neurologists utilize electroencephalography (EEG) to record epileptic seizures. Automated seizure identification is a crucial step in the diagnosis of epilepsy overcoming the limitations of visual diagnosis. This study makes use of the Children’s Hospital of the Massachusetts Institute of Technology, Boston (CHB-MIT) dataset. The seizure detection system implemented in this research utilized the variational mode decomposition (VMD) method and Light Gradient Boosting Classifier (LGBM). The seizure and no seizure signals are separated using the K-fold cross-validating technique after the linear and non-linear properties are recovered from the band limited-intrinsic mode functions (BL-IMFs). This approach performance is evaluated with the sensitivity, specificity, and accuracy metrics and achieved an average of 82.32%, 86.49%, and 90.53% respectively for all patients in the dataset.