Improved ensemble learning model with optimal feature selection for automated epileptic seizure detection

his work plans to propose the novel epileptic seizure detection using the Improved Ensemble Learning Model (I-ELM). The proposed model focuses five different stages Pre-processing, signal decomposition, Feature extraction, optimal feature selection, and detection. In the initial process, the EEG signals are gathered from two public sources, and pre-processing of the signals is performed by the Short-Time Fourier Transform (STFT) technique. Then, the Discrete Wavelet Transform (DWT) is performed for the signal decomposition. Feature extraction technique is enforced to the decomposed signals using the temporal and spatial feature extraction techniques. To reduce the feature-length for managing the training complexity and enhancing detection performance, a new meta-heuristic algorithm called Modified Tunicate Swarm Algorithm (M-TSA) is adopted for accurate feature selection. These optimally selected features are used for the detection of I-ELM, which utilise three main classifiers like Fuzzy classifier, Long Short Term Memory (LSTM) and Deep Neural Network (DNN). I-ELM model is accomplished based on the high ranking from the heuristically improved ensemble classifiers using the proposed M-TSA, which detects the seizure from the EEG signals. From the experimental outputs based on different EEG datasets are compared with the traditional models and revealed the efficiency of the suggested model for identifying epileptic EEGs, denoting its powerful capacity in automated seizure detection.