Algorithm 1: Pseudo code for computing the AUC-PR based on the continuous interpolation.

Initially, we choose the classification threshold such that the number of true positives is equal to the total number of positives. Then we iterate as long as the number of true positives – and, hence, recall – is greater than . We determine the new point by choosing the next existing score as classification threshold. Unless this threshold leads to an identical number of true positives, we compute the values of , , and as defined by equation (6), and set the borders of the integration. We use these values to compute the AUC between the current points and , and proceed with the while-loop. After termination of the loop, holds the AUC-PR.