Earthquakes are natural disasters with the potential for catastrophic destruction and loss of life. This study aims to enhance earthquake prediction accuracy, focusing on earthquake magnitude and likelihood through Machine Learning (ML) models trained on historical seismic data. Using the earthquake dataset, which contains data on earthquake events from 1966 to 2007, we apply four ML models: linear regression, Support Vector Machine (SVM), Naive Bayes, and Random Forest. Each model is trained to identify patterns by analyzing key earthquake parameters such as magnitude, location, depth, and seismic station data which are known to influence seismic event characteristics. We evaluate predictive accuracy using Mean Squared Error (MSE) and R² scores to determine the most effective model. By comparing these performance metrics, we identify which model performs best in accurately predicting earthquake magnitudes and identifying potential future occurrences. Initial results indicate that ensemble methods, such as Random Forest, tend to outperform simpler models due to their ability to capture complex feature interactions. Our findings underscore the importance of model choice in earthquake prediction and suggest that integrating more data and real-time monitoring can substantially enhance prediction accuracy. This study highlights the potential for machine learning to contribute to more reliable earthquake prediction systems, with the long-term goal of improving public safety and readiness in earthquake-prone areas. By demonstrating that machine learning models can leverage historical earthquake data for predictive purposes, we suggest a pathway toward implementing more advanced, data-driven forecasting model, which could ultimately support early warning systems and disaster preparedness efforts.