Why are the example predictions not one of the 5 discrete values (0, 0,25, 0.5, 0.75, 1) ? Would my predictions suffer if I submitted them that way?
I suspect there is something obvious I’m missing.
Yeah, it’s a bit strange, but your predictions should be real-valued continuous like the example file, and not just the 5 values in the targets. You are not scored on error, but on ranking correlation, so the first thing that happens is your predictions are ranked and any ties eliminated. If you submit only with those 5 values, you may suffer, you may gain, but there will be a big random component to your scores so you probably don’t want that…
4 Likes
It is often the case that a number is naturally associated to the outcome of a random experiment: the number of boys in a three-child family, the number of defective light bulbs in a case of 100 bulbs, the length of time until the next customer arrives at the drive-through window at a bank. Such a number varies from trial to trial of the corresponding experiment, and does so in a way that cannot be predicted with certainty; hence, it is called a random variable . In this chapter and the next we study such variables.
Never knew these existed!