This work performs a comprehensive review of the different machine learning techniques applied in earthquake magnitude prediction. Earthquakes pose serious threats to life and infrastructure; thus, the accurate prediction of earthquakes becomes one of the prime needs for early warning and disaster management. Traditional machine learning models such as Random Forest and Support Vector Machines are widely used but most often rely on handcrafted features. In contrast, deep learning models, including Convolutional Neural Networks and Long Short-Term Memory networks, have substantially outperformed the techniques so far, by modeling complex nonlinear relationships from seismic data. While promising results are obtained, several challenges remain regarding dealing with scarce data, class imbalance, and intrinsic complexity of the earthquake phenomena. Some of the future research directions espoused in the paper involve advanced deep learning architecture integrated with AI and IoT and semi-supervised learning, where the aim is to improve the accuracy of prediction and facilitate real-time forecasting systems to better equip disaster preparedness.