import os
import sys
import time
import numpy
import theano
import theano.tensor as T
from theano.tensor.shared_randomstreams import RandomStreams
from pprint import pprint
from scipy.io import matlab
import re
import math
from theano import shared
from collections import OrderedDict
#from my_pytable_FS_sulfur import *
import random

Method = sys.argv[1]
thres = float(sys.argv[2])


if __name__ == '__main__':


	all_pos_prob= numpy.load("../results/prob_scores/"+Method+"/prob_y.dat")
	all_y_true= numpy.load("../results/prob_scores/"+Method+"/true_y.dat")

	if Method == 'FEATURE_SVM' or Method == 'Voxel_SVM':
		all_pos_prob=all_pos_prob[:,1]


	tp_total = 0
	fp_total = 0
	tn_total = 0
	fn_total = 0

	pred_pos = numpy.where(all_pos_prob>=thres)
	pred_neg = numpy.where(all_pos_prob<thres)

	all_y_pred = numpy.zeros(all_y_true.shape)

	for i in range (0,len(pred_pos)):
		all_y_pred[pred_pos[i]]=1


	from sklearn.metrics import confusion_matrix
	m=confusion_matrix(all_y_true,all_y_pred)
	print m
	tp = m[1][1]
	fp = m[0][1]
	tn = m[0][0]
	fn = m[1][0]
	tp_total += tp
	fp_total += fp
	tn_total += tn
	fn_total += fn
	n=numpy.zeros((m.shape[0],m.shape[1]))

	print "tp_total: "+str(tp_total)
	print "fp_total: "+str(fp_total)
	print "tn_total: "+str(tn_total)
	print "fn_total: "+str(fn_total)


	precision=float(tp_total)/(tp_total+fp_total)
	recall=float(tp_total)/(tp_total+fn_total)
	acc = float(tp_total+tn_total)/(tp_total+tn_total+fp_total+fn_total)
	specificity = float(tn_total)/(tn_total+fp_total)
	MCC_up = ((tp_total*tn_total)-(fn_total*fp_total))
	MCC_bottom = numpy.sqrt((tp_total+fn_total)*(tp_total+fp_total)*(tn_total+fn_total)*(tn_total+fp_total))
	MCC = float(MCC_up) / MCC_bottom
	print "precision:"+str(precision)
	print "recall:"+str(recall)
	print "specificity:"+str(specificity)
	print "acc:"+str(acc)
	print "MCC:"+str(MCC)


	from sklearn import metrics
	from sklearn.metrics import roc_auc_score
   
	fpr, tpr, thresholds = metrics.roc_curve(all_y_true,all_pos_prob)
	S_1 = metrics.auc(fpr, tpr)
	S_2 = roc_auc_score(all_y_true,all_pos_prob)

	print ("AUC score:"+str(S_1))

	import numpy as np
		
	point5_pc_low = np.where(0.5/100<=fpr) 
	point5_pc_high =np.where(fpr<=0.5/100)
	last=point5_pc_high[0][-1]
	last=point5_pc_low[0][0]
	
	ER_5pt=tpr[last]/fpr[last]

	one_pc_low = np.where(1.0/100<=fpr) 
	one_pc_high =np.where(fpr<=1.0/100)

	last=one_pc_high[0][-1]
	last=one_pc_low[0][0]
	ER_one=tpr[last]/fpr[last]

	two_pc_low = np.where(2.0/100<=fpr) 
	two_pc_high =np.where(fpr<=2.0/100)
	last=two_pc_high[0][-1]
	last=two_pc_low[0][0]
	
	ER_two=tpr[last]/fpr[last]

	
	five_pc_low = np.where(5.0/100<=fpr) 
	five_pc_high = np.where(fpr<=5.0/100)

	last=five_pc_high[0][-1]
	last=five_pc_low[0][0]
	
	ER_five=tpr[last]/fpr[last]

	print "RE 0.5:"
	print ER_5pt

	print "RE 1:"
	print ER_one

	print "RE 2:"
	print ER_two

	print "RE 5:"
	print ER_five