In [1]:

    

import random
import pandas
import numpy as np
import tensorflow as tf
from sklearn import metrics, cross_validation
from tensorflow.contrib import skflow


    

In [2]:

    

random.seed(42)


    

In [3]:

    

data = pandas.read_csv('data/titanic_train.csv')


    

In [4]:

    

X = data[["Embarked"]]


    

In [5]:

    

y = data[["Survived"]]


    

In [6]:

    

X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.2, random_state=42)


    

In [7]:

    

embarked_classes = X_train["Embarked"].unique()


    

In [8]:

    

print('Embarked has next classes: ', embarked_classes)


    

    




Embarked has next classes:  ['S' 'C' 'Q' nan]

In [9]:

    

cat_processor = skflow.preprocessing.CategoricalProcessor()


    

In [10]:

    

X_train = np.array(list(cat_processor.fit_transform(X_train)))


    

In [11]:

    

X_test = np.array(list(cat_processor.fit_transform(X_test)))


    

In [12]:

    

n_classes = len(cat_processor.vocabularies_[0])


    

In [13]:

    

EMBEDDING_SIZE = 3


    

In [14]:

    

def categorical_model(X, y):
    features = skflow.ops.categorical_variable(
        X, n_classes, embedding_size=EMBEDDING_SIZE, name='embarked')
    return skflow.models.logistic_regression(tf.squeeze(features, [1]), y)


    

In [15]:

    

classifier = skflow.TensorFlowEstimator(model_fn=categorical_model, n_classes=2)


    

In [16]:

    

classifier.fit(X_train, y_train['Survived'])


    

    




Step #100, epoch #4, avg. train loss: 0.65760
Step #200, epoch #8, avg. train loss: 0.65196






    
Out[16]:





TensorFlowEstimator(batch_size=32, class_weight=None, clip_gradients=5.0,
          config=None, continue_training=False, learning_rate=0.1,
          model_fn=<function categorical_model at 0x1127420d0>,
          n_classes=2, optimizer='Adagrad', steps=200, verbose=1)

In [17]:

    

print("Accuracy: {0}".format(metrics.accuracy_score(classifier.predict(X_test), y_test)))


    

    




Accuracy: 0.6256983240223464

In [18]:

    

print("ROC: {0}".format(metrics.roc_auc_score(classifier.predict(X_test), y_test)))


    

    




ROC: 0.6105506155950753

In [19]:

    

def one_hot_categorical_model(X, y):
    features = skflow.ops.one_hot_matrix(X, n_classes)
    return skflow.models.logistic_regression(tf.squeeze(features, [1]), y)


    

In [20]:

    

classifier = skflow.TensorFlowEstimator(model_fn=one_hot_categorical_model,
                                       n_classes=2, steps=1000, learning_rate=0.01)


    

In [21]:

    

classifier.fit(X_train, y_train['Survived'])


    

    




Step #100, epoch #4, avg. train loss: 0.84168
Step #200, epoch #8, avg. train loss: 0.75244
Step #300, epoch #13, avg. train loss: 0.72017
Step #400, epoch #17, avg. train loss: 0.69646
Step #500, epoch #21, avg. train loss: 0.68045
Step #600, epoch #26, avg. train loss: 0.67679
Step #700, epoch #30, avg. train loss: 0.67083
Step #800, epoch #34, avg. train loss: 0.66649
Step #900, epoch #39, avg. train loss: 0.66256
Step #1000, epoch #43, avg. train loss: 0.66194






    
Out[21]:





TensorFlowEstimator(batch_size=32, class_weight=None, clip_gradients=5.0,
          config=None, continue_training=False, learning_rate=0.01,
          model_fn=<function one_hot_categorical_model at 0x1129a6158>,
          n_classes=2, optimizer='Adagrad', steps=1000, verbose=1)

In [22]:

    

print("Accuracy: {0}".format(metrics.accuracy_score(classifier.predict(X_test), y_test)))


    

    




Accuracy: 0.6201117318435754

In [23]:

    

print("ROC: {0}".format(metrics.roc_auc_score(classifier.predict(X_test), y_test)))


    

    




ROC: 0.6027310924369748

In [ ]: