Earthquake prediction plays a vital role in minimizing damage and saving lives by providing early warnings. Generally, two main categories of earthquake prediction exist: short-term predictions, made hours or days in advance, and forecast predictions,made months or years ahead. Most existing studies focus on long-term forecasting by analyzing the historical seismic activity of specific regions. This work aims to classify earthquake events as major (positive) or non-major (negative) using advanced machine learning algorithms. A real-world earthquake dataset has been utilized to train and test eight different models: Random Forest, Naive Bayes, Logistic Regression, Multi-Layer Perceptron, AdaBoost, K-Nearest Neighbors (KNN), and Classification and Regression Trees (CART). For each model, various hyperparameters were tuned to enhance performance. Comparative analysis was conducted using standard evaluation metrics, and the results demonstrate that three algorithms achieved highly reliable prediction accuracy for identifying major seismic events. Index Terms Machine Learning, Earthquake Prediction, Seismic Forecasting, Natural Disaster Analysis