Learn some of the main baselines used for time series data and how to calculate them.


Baselines are important so that we can have something to compare our models to. Imagine we put a lot of work into creating an LSTMLong Short-Term Memory neural network model to forecast sales, and the model doesn't perform better than just using a simple average as the forecast.

So, to evaluate our model's performance, we compare it to a benchmark, which is our baseline. If our model does worse than this, it's probably not a good model.

Simple average

As the name suggests, simple average means just forecasting the mean value for every point in time. If our data has no trend, this might be a good baseline. The formula to calculate the mean is:

Get hands-on with 1200+ tech skills courses.