Earthquakes are both highly deadly and often unforeseen. With not many factors indicating the occurrence of an earthquake, and of those factors that are prominent indicators, many are unreliable. These can include the presence of radon gas or even unusual activity from animals, both of which can occur without the event of an earthquake following. This project aims to predict the magnitudes of earthquakes prior to their occurrence based on previous earthquakes within that area by taking advantage of resources such as machine learning. The model designed took in input of about 100,000 data points, provided by USGS, and was clustered into 10 clusters using K-Means. These clusters were then treated as datasets within the random forest regression model. The final results comprised -0.031, 0.439, 0.193, and 0.329 mean residual, root mean squared error, mean squared error, and mean absolute error, respectively, illustrating the overall slight underestimation of the model’s predictions. Additionally, the random forest model typically underestimated the value of the actual magnitude. At the same time, the root mean squared error value was somewhat low, highlighting the efficiency and accuracy of a random forest model for regression. The research used this random forest model to use latitude and longitude for current and previous data to predict whether an earthquake large enough will impact society and call for evacuation.