plot_results.py

"""
Plot results of the criterion
Author : Diviyan Kalainathan
Date : 11/10/2016
"""
import os
import pandas as pd
import matplotlib.pyplot as plt
import cPickle as pickle
import numpy
from itertools import cycle
from sklearn.metrics import auc, average_precision_score,precision_recall_curve
import seaborn as sns
from matplotlib import rc
rc('font',**{'family':'sans-serif','sans-serif':['Helvetica']})
## for Palatino and other serif fonts use:
#rc('font',**{'family':'serif','serif':['Palatino']})
rc('text', usetex=True)

inputdata='output/test/'
colors = cycle(['cyan', 'indigo', 'seagreen', 'gold', 'blue',
                'darkorange','red','grey','darkviolet','mediumslateblue','chocolate'])
crit_names_label = [#"Pearson's correlation",
              #"AbsPearson's correlation",
              #"Pvalue Pearson",
              r"$\textrm{Chi2 test}$",
              r"$\textrm{NMutual information}$",
              r"$\textrm{AMutual information}$",
              r"$\textrm{Corrected Cramer's V}$",
              r"$\textrm{Mercer Kernels}$",
               #"Lopez-Paz Causation coefficient",
              r"$\textrm{FSIC}$"
               #"BF2d mutual info",
               #"BFMat mutual info",
               #"ScPearson correlation",
               #"ScPval-Pearson"
              ]
crit_names = [#"Pearson's correlation",
              #"AbsPearson's correlation",
              #"Pvalue Pearson",
              "Chi2 test",
              "NMutual information",
              "AMutual information",
              "Corrected Cramer's V",
              "Mercer Kernels",
               #"Lopez-Paz Causation coefficient",
              "FSIC"
               #"BF2d mutual info",
               #"BFMat mutual info",
               #"ScPearson correlation",
               #"ScPval-Pearson"
              ]
results=[]
f, axarr = plt.subplots(2, sharex=True)
fig1 = plt.figure()
ax1 = fig1.add_subplot(111)
fig2=plt.figure()
ax2=fig2.add_subplot(111)

# for crit in crit_names:
#     inputfile=inputdata+'test_crit_l_'+crit[:4]+'-1.csv'
#     if os.path.exists(inputfile):
#         print(crit),
#         df=pd.read_csv(inputfile,sep=';')
#         df=df[['Target','Pairtype']]
#         df=df[numpy.isfinite(df['Target'])]
#         #df['Target']=df['Target'].astype('float')
#         #print(df.dtypes)
#         if crit[:4]!='Lope':
#             df=df.sort_values(by='Target', ascending=False)
#         else:
#             df=df.sort_values(by='Target', ascending=True)

#         #print(df)
#         N_max=len(df.index)
#         print(N_max),
#         N=0.0
#         Mprobes_max = (df['Pairtype']=='P').sum()
#         print(Mprobes_max)
#         Mprobes = 0.0
#         FDR=[]

#         for index,row in df.iterrows():
#            N=N+1
#            if row['Pairtype']=='P':
#                Mprobes+=1

#            FDR.append((N_max/N)*(Mprobes/Mprobes_max))

#         results.append(FDR)


# pickle.dump(results,open(inputdata+'res.p','wb'))
# #results=pickle.load(open(inputdata+'res.p','rb'))

# '''for i in range(len(results)-1):
#     print(results[i]==results[-1])

# print(len())'''
# #print(results)
# for i,color in zip(results,colors):
#     ax1.plot(range(len(i)),i,color=color)
#     ax2.plot(range(len(i)),i,color=color)
# plt.legend(crit_names_label,loc=4)
# plt.xlabel(r"$\textrm{Number of probes retrieved}$")
# plt.ylabel(r"$\textrm{False discovery rate}$")
# ax1.set_xscale('log')
# plt.show()#FDR w/ probes

colors = cycle(['cyan', 'indigo', 'seagreen', 'gold', 'blue', 'darkorange','red','grey','darkviolet','mediumslateblue'])
for crit, color, name_label in zip (crit_names,colors,crit_names_label):
    tpr=[]
    fpr=[]
    ppv=[]


    print(crit)
    try:
        with open("kaggle/norm_fd/CEfinal_train_"+crit[:4]+'.csv','r') as results_file:
            df=pd.read_csv(results_file,sep=';')
            df=df.sort_values(by='Target', ascending=False)

            P=float((df['Pairtype']!=4).sum())
            Plist=(df['Pairtype']!=4).tolist()
            N=float((df['Pairtype']==4).sum())

            TP=0.0
            FP=0.0
            for index,row in df.iterrows():
                if crit[:4]!='Lope':
                    if row['Pairtype']==4:
                        FP+=1.0
                    else:
                        TP+=1.0
                else:
                     if row['Pairtype']!=4:
                        FP+=1.0
                     else:
                        TP+=1.0

                tpr.append(TP/P) #TPR=recall
                fpr.append(FP/N) #FPR
                ppv.append(TP/(TP+FP))
            tpr,fpr, ppv= (list(t) for t in zip(*sorted(zip(tpr,fpr,ppv))))
            auc_score=auc(fpr,tpr)
            pres,rec,_= precision_recall_curve(Plist,df['Target'].tolist())
            ac_pr_score=average_precision_score(Plist,df['Target'].tolist())
            pl1=ax1.plot(fpr,tpr,label=name_label+r' $ (\textrm{area} : $'+r' ${0:3f})$'.format(auc_score),color=color)
            pl2=ax2.plot(rec,pres,label=name_label+r' $ (\textrm{area} : $'+r' ${0:3f})$'.format(ac_pr_score),color=color)

    except IOError:
        continue


ax1.plot([0, 1], [0, 1], linestyle='--', color='k',
         label=r"$\textrm{Luck}$")

ax1.set_xlabel(r"$\textrm{False Positive Rate}$")
ax1.set_ylabel(r"$\textrm{'True Positive Rate}$")
ax1.set_title(r"$\textrm{ROC Curve of independence tests on the}$"+" \n"+r"$\textrm{ Chalearn Cause-effect pairs challenge dataset}$")
ax1.legend(loc="lower right")
ax2.set_xlabel(r"$\textrm{Recall}$")
ax2.set_ylabel(r"$\textrm{Precision}$")
ax2.set_title(r"$\textrm{Precision recall curves on the}$"+ "\n" +r"$\textrm{Chalearn Cause-effect pairs challenge dataset}$")
ax2.legend(loc='best')
plt.show()