Linear Regression is a simple machine learning model for regression problems, i.e., when the target variable is a real value.
Good question, I think it refers to the cost of algorithm error, in other words the error that will be incurred to use one model or another