# Time series Prediction: Coverage vs Accuracy

I'm building an algorithm which looks at a sequence of numbers and produces a prediction.

This prediction is then compared against real life data, and it should say if it was correct or not.

The question is:

Set X = {1,2,3,4,5,6,7,8,9,10,11,12}, the numbers to predict Set Y = {1,2,5,6}, the predicted numbers

How are the coverage and accuracy calculated?

My thinking was coverage means 30%, because 4 out of 12 numbers are predicted correctly.

For accuracy, I can't figure it out, but I found this text in a paper:

The accuracy of the algorithm is defined as follows.
For a given sequence S=(e1,...,ek,ek+1) we say that
A is correct on S if A(e1,...,ek) = ek+1.
For a (multi)set Y of sequence, the accuracy of A on
Y is the fraction of sequences in Y for which A is
correct. We seek an algorithm which provides a high
accuracy.


Any feedback would be highly appreciated.