The Bayes factor is a metric used to decide between two competing models that describe a process: in our case, a predictive algorithm. I demonstrate how a Bayes factor can be used to decide whether a sophisticated algorithm outperforms a Naive algorithm that simply predicts the most common label.
Let M1 and M2 be two competing models. M1 is likely to be the model we are considering replacing M2, and a larger K indicates more evidence that M1 is the better model. If M1 denotes our predictive algorithm, a large K indicates that our algorithm does a good job of predicting the target variable.
Now recall that, in Bayesian statistics, we have a prior distribution, we collect data, and we compute a posterior distribution for an event. ...
