Linear regression is a statistical method used in machine learning that models the relationship between a dependent variable and one or more independent variables by fitting a linear equation to observed data. The main goal is to find a linear relationship that can be used to predict the values of the dependent variable based on the values of the independent variables.