We must know the way sure our model is about its predictions to make well-informed decisions. Hence, returning only a degree prediction isn’t enough. It doesn’t tell us anything about whether we will trust our model or not. If you need to know why, try my article below.
Within the article, I exploit a classification problem for instance. Nevertheless, many real-world problems are regression problems. For instance, we would like to know the way certain a model is when predicting tomorrow’s temperature.
Because the temperature is a continuous variable, we would love to know during which interval the true temperature will lie.
The broader the interval, the more uncertain the model. Hence, we should always trust it less when making decisions.
Two approaches come to mind. Either we use a set of models that predict the interval or we turn a degree prediction right into a…
