One straightforward measure of accuracy is the Matthews correlation coefficient C that can be calculated from the number of true positives (tp), true negatives(tn), false positives(fp) and false negatives(fn) described previously and also shown graphically on the previous figure. These numbers can be counted for any given threshold value that separates the two groups.


The value of C is close to 1 if the members of the group close mostly all above the threshold value, while the non-members score below. It is around zero, if the two groups score about equally on both sides of the threshold value (no separation). 2