from matplotlib import pyplot as plt

import sklearn

from sklearn.metrics import roc_curve, auc

from sklearn.model_selection import train_test_split

from sklearn.linear_model import LogisticRegression

from sklearn.tree import DecisionTreeClassifier

from sklearn.ensemble import RandomForestClassifier

from sklearn.naive_bayes import GaussianNB

import pandas as pd

import sklearn.datasets

def get_iris_df():

return df

df = get_iris_df()

CLASS_MAP = {'Logistic Regression': ('-', LogisticRegression()), 'Naive Bayes':('--', GaussianNB()), 'Desicion Tree': ('.-', DecisionTreeClassifier(max_depth=5)), 'Random Forest': (':', RandomForestClassifier(max_depth=5, n_estimators=10, max_features=1)),}

X, Y = df[df.columns[:3]], (df['species']=='virginica')

X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=.8)

for name, (line_fmt, model) in CLASS_MAP.items():

model.fit(X_train, Y_train)

preds = model.predict_proba(X_test)

pred = pd.Series(preds[:,1])

fpr, tpr, thresholds = roc_curve(Y_test, pred)

auc_score = auc(fpr, tpr)

label = '%s: auc=%f' % (name, auc_score)

plt.plot(fpr, tpr, line_fmt, linewidth=5, label=label)

plt.legend(loc="lower right")

plt.title("Comparisons among classifiers")

plt.plot([0,1], [0,1], 'k--')

plt.xlim([0.0, 1.0])

plt.ylim([0.0, 1.05])

plt.xlabel('False Positive Ratio')

plt.ylabel('True Positive Ratio')

plt.show()