% matplotlib inline
import itertools
import matplotlib.pyplot as plt
import os
import pandas as pd
import seaborn as sns
from sklearn import cluster
from sklearn import cross_validation
from sklearn import datasets
from sklearn import ensemble
from sklearn import linear_model
from sklearn import metrics
from sklearn import naive_bayes
from sklearn import preprocessing
from sklearn import svm
from sklearn import tree

plt.style.use('fivethirtyeight')

scoring_funcs = [ii for ii in dir(metrics) if 'score' in ii and not ii.startswith('_')]

_DATA_DIR = 'data'
_TRAINING_DATA_PATH = os.path.join(_DATA_DIR, 'userart-mat-training.csv')
_TEST_DATA_PATH = os.path.join(_DATA_DIR, 'userart-mat-test.csv')
_ARTIST_TAGS_PATH = os.path.join(_DATA_DIR, 'artists-tags.txt')

df = pd.read_csv(_TRAINING_DATA_PATH)

del df['ArtistID']

n_artists, n_users = X.shape
frame_number = n_artists // n_users + 1

plt.figure(figsize=(16, 12))
plt.plot(n_clusters, silhouette_scores)
plt.ylabel("Silhouette Score");
plt.xlabel("Cluster");
plt.title("Silhouette vs Cluster Size");

iris = datasets.load_iris()
X = iris.data
y = iris.target

plt.figure(figsize=(16, 12))
plt.imshow(X.reshape((30, 20)), cmap='bone')
plt.xticks([]);
plt.yticks([]);

n_clusters = range(2, 15)
adjusted_random_scores = []
v_measure_scores = []

for n_cluster in n_clusters:
    print('K-Means run for number of clusters: {} is started'.format(n_cluster))
    kmeans = cluster.KMeans(n_clusters=n_cluster, init="k-means++").fit(X)
    labels = kmeans.labels_
    adjusted_random_scores.append(metrics.adjusted_rand_score(y, labels) )
    v_measure_scores.append(metrics.v_measure_score(y, labels))

plt.figure(figsize=(16, 12))
plt.plot(n_clusters, adjusted_random_scores)
plt.ylabel("Adjusted Random Score");
plt.xlabel("Cluster");
plt.title("Adjusted Random Score vs Cluster Size");

plt.figure(figsize=(16, 12))
plt.plot(n_clusters, v_measure_scores)
plt.ylabel("V Measure Score");
plt.xlabel("Cluster");
plt.title("V Measure Score vs Cluster Size");

df = sns.load_dataset("titanic")
df.head(10)

df.dropna(inplace=True)

label_encoder = preprocessing.LabelEncoder()
df['embarked'] = label_encoder.fit_transform(df.embarked)
df['embark_town'] = label_encoder.fit_transform(df.embark_town)
df['alive'] = label_encoder.fit_transform(df.alive)
df['alone'] = label_encoder.fit_transform(df.alone)
df['deck'] = label_encoder.fit_transform(df.deck)
df['adult_male'] = label_encoder.fit_transform(df.adult_male)
df['who'] = label_encoder.fit_transform(df.who)
df['class'] = label_encoder.fit_transform(df['class'])
df['parch'] = label_encoder.fit_transform(df['parch'])
df['sex'] = label_encoder.fit_transform(df.sex)

y = df.survived.as_matrix()
del df['survived']

X = df.as_matrix().astype(float)

X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=.5,
                                                    random_state=2)

classifier_name_tuples = [('Sigmoid Kernel Support Vector Classifier', svm.SVC(kernel='sigmoid', probability=True, shrinking=True, gamma=.1)),
                          ('Bernoulli Naive Bayes Classifier', naive_bayes.BernoulliNB()),
                          ('RBF Kernel Support Vector Classifier', svm.SVC(kernel='rbf', probability=True)),
                          ('Extra Trees Classifier', ensemble.ExtraTreesClassifier()),
                          ('Logistic Regression', linear_model.LogisticRegression(penalty='l1', C=.1)),
                          ('Random Forest Classifier', ensemble.RandomForestClassifier()),
                          ('Multinomial Naive Bayes Classifier', naive_bayes.MultinomialNB())
                          ]

    
colors = itertools.cycle(["#30a2da",
    "#fc4f30",
    "#e5ae38",
    "#6d904f",
    "#8b8b8b",
    "dodgerblue",
    "mediumpurple",
])

fig, ax = plt.subplots(figsize=(16,12))
for clf_name, clf in classifier_name_tuples:
    clf.fit(X_train, y_train)
    probas_ = clf.predict_proba(X_test)
    fpr, tpr, thresholds = metrics.roc_curve(y_test, probas_[:, 1], )
    roc_auc = metrics.auc(fpr, tpr)
    plt.plot(fpr, tpr, color=colors.next(), lw=2, label='ROC curve for {}'.format(clf_name))
    print('Region under curve for {0} is : {1:.2f}'.format(clf_name, roc_auc))
#    ax.annotate('AUC for {0} = {1:.2f}'.format(clf_name, roc_auc), xy=(0.35, 0.6));

plt.plot([0, 1], [0, 1], color='black', lw=2, linestyle='dotted', label='random guess')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.0])
plt.xlabel('False Positive Rate');
plt.ylabel('True Positive Rate');
plt.title('ROC curves for different classifiers');
plt.legend(loc="lower right");