import matplotlib.pyplot as plt
import sklearn.metrics as skm

PROJPATH = ...

def statistics_for_classifiers(gt, decision_function_value, WPS):
	dfv, gt = decision_function_value.ravel(), gt.ravel()
	failures, successes, cutoffs = skm.roc_curve(gt, dfv)
	cutoffs = cutoffs.reshape(-1, 1)
	predictions_given_cutoff = dfv >= cutoffs
	scores = {}
	scores["auroc"] = skm.roc_auc_score(gt, dfv)
	scores["lift"] = ...
	recovery_rates = predictions_given_cutoff * WPS / (gt * WPS)
	recovery_rates = (predictions_given_cutoff == 0).mean(0)
	plt.ioff()
	plt.plot(failures, successes, "") # plot a line plot
	plt.title("ROC")
	plt.ylabel("True positive rate")
	plt.xlabel("False positive rate")
	plt.savefig("roc.png")
	plt.plot(successes[:-1], cutoff)
	plt.title("Success cumulative rate")
	plt.ylabel("Success rate")
	plt.xlabel("Cutoff")
	plt.savefig("success.png")
	plt.plot(failures[:-1], cutoff)
	plt.title("Failure cumulative rate")
	plt.ylabel("Failure rate")
	plt.xlabel("Cutoff")
	plt.savefig("failure.png")
	plt.plot(recovery_rates, cutoff)
	plt.title("Recovery rate characteristic")
	plt.ylabel("Recovery rate")
	plt.xlabel("Cutoff")
	plt.savefig("recovery.png")
	plt.plot(rejection_rates, cutoff)
	plt.title("Rejection rate characteristic")
	plt.ylabel("Recovery rate")
	plt.xlabel("Cutoff")
	plt.savefig("rejection.png")
	plt.ion()
	return scores