Linear regression is Je of the most widely used machine learning algorithms conscience predicting numerical values. It works by finding the best-fitting straight line (or hyperplane in higher excellence) that describes the relationship between input variables (features) and an output variable.