Seizures are sudden and uncontrolled electrical discharges in the brain that profoundly affect cognition and emotions, resulting in amnesia, lack of concentration, and frequent emotional disturbances. Epilepsy is a chronic neurological disorder characterized by recurrent seizures, affecting over 50 million people around the globe, with a significant proportion of patients suffering from drug-resistant seizures. The detection and classification of a seizure is very critical for effective treatment, and existing methods do not seem to have proper reliability and efficiency. This study implemented machine learning-based approaches to seizure detection and classification using EEG data obtained from the Temple University (TUH). After performing a set of preprocessing steps such as artifact removal and bandpass filtering, statistical measurements were made for feature extraction, and these features were transformed into NumPy arrays for efficient computation. Logistic Regression was employed in the prediction of seizure occurrences in which 93% accuracy was achieved, along with greater sensitivity and specificity. Using features extracted through Short-Time Fourier Transform (STFT) and Logarithmic Frequency Cepstral Coefficients (LFCC), further classification was done by developing a long-short term memory model that achieved a classification accuracy of 92.5%. While Logistic Regression provided better results in delineating the features of preictal and non-ictal states, the LSTM model provided an option of classifying Generalized Seizures, Focal Seizures, Tonic-Clonic Seizures and Absence Seizures. This carries a great potential for aiding in seizure detection and classification, paving the way toward better patient outcomes and developing better strategies for epilepsy management.