Now, if you knew the underlying distribution, you could very simply figure out the probability that the person was at risk given x and the probability that the person was not at risk given x (where x refers to the vector for a single person or an instance of X). P (y = at-risk | x) and P (y = not-at-risk | x) Then, one way to make a prediction would be to choose the category that had the higher probability. Later in the chapter, we’ll come to just how you can do this (it involves using Bayes’s theorem), but for now, all we need to appreciate is that this is the best an ML algorithm can do,
...more
