Generates decision values from a binormal distribution and plots true vs. empirical vs. model-based quantities derived from the confusion matrix of the simulated classifier.
[targs, dvs] = prc_generate_dvs(n, alpha, muN, sigmaN, muP, sigmaP)
n: number of examples
alpha: fraction of examples from the positive class (range: 0..1)
- 0 = total class imbalance in favour of the negative class
- 0.5 = no class imbalance
- 1 = total class imbalance in favour of the positive class
muN, sigmaN: mean and std.dev. of decision values assigned to examples from negative class
muP, sigmaP: mean and std.dev. of decision values assigned to examples from positive class
K.H. Brodersen, C.S. Ong, K.E. Stephan, J.M. Buhmann (2010). The binormal assumption on precision-recall curves. In: Proceedings of the 20th International Conference on Pattern Recognition (ICPR).