An Earthquake Early Warning (EEW) system works to identify seismic activity and issue alerts before the earth- shattering reaches the affected areas, possibly rescuing lives and minimizing damage. Support Vector Machines (SVM), a type of SVM algorithm, can be leveraged effectively in creating such systems. EEW systems play a crucial role in reducing the impact of earthquakes on people and buildings. In this paper, a new method has been introduced using machine learning techniques, specifically SVM and RF, to quickly and accurately find where an earthquake starts in EEW systems. SVMs are really best at classifying data, whether it’s linear or non-linear. They do this by finding a hyperplane that can amazing in separate different classes of data points. When the data is non-linear, SVMs use kernel functions to convert the data into a higher-dimensional space where it becomes possible to separate it linearly. We collected the dataset from the USGS earthquake website. Dataset consists of streaming data from the USGS website. We use information about when earthquake waves first reach five recording stations, comparing them to a main station to pinpoint the epicenter. Testing our model with earthquake data from Japan, as we achieved the following results in which a Mean Absolute Error (MAE) of 2.88 km, showing high precision. Even with only 10% of the data and three recording stations, our system still performs well (MAE < 5 km). This new approach promises better accuracy, adaptability, and speed in predicting earthquake locations for EEW.