*Linear Regression* is a great algorithm to use as a baseline when working on a new problem. It is simple to evaluate and understand. Training such a model can be extremely fast for vast amounts of data. You can often be surprised at how good this simple algorithm can do.

Of course, this algorithm can perform very poorly under certain circumstances. One of the primary (practical) disadvantages of it is that it can't separate non-linear data. Your predictions assume some linear relationship.