A Support Vector Machine (SVM) is a type of supervised learning algorithm used in machine learning for classification and regression tasks. It works by finding the hyperplane that best divides a dataset into classes with the maximum margin between them.