-
Notifications
You must be signed in to change notification settings - Fork 2
/
MortgageDeepLearning.py
252 lines (203 loc) · 9.26 KB
/
MortgageDeepLearning.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
from __future__ import absolute_import
from __future__ import print_function
from datetime import datetime
from PIL import Image
from keras.layers.normalization import BatchNormalization
from keras.models import Sequential
from keras.layers.core import AutoEncoder, Dense, Activation, Dropout
from keras.layers import containers
from keras.utils import np_utils
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
from sklearn.cross_validation import train_test_split
from sklearn.decomposition import TruncatedSVD
from sklearn.manifold import TSNE
from sklearn.metrics import classification_report
from sklearn.preprocessing import StandardScaler, Imputer
import seaborn as sns
sns.set(style='white', palette='Set2')
from DataDescription import getDefaultData, getNonDefaultData
from utils import tile_raster_images
def plotExamples(data, number = 9, title=""):
title = title + "Example_data"
print("Plotting " + title)
image_size = np.floor(np.sqrt(data.shape[1]))
subplot_size = int(np.floor(np.sqrt(number)))
raster = tile_raster_images(
X=data,
img_shape=(image_size, image_size), tile_shape=(subplot_size, subplot_size),
tile_spacing=(1, 1))
image = Image.fromarray(raster)
image.save("plots/" + datetime.now().strftime("%Y%m%dT%H%M%S") + "_" + title + '.png')
def getActivationsLayer0(layers, requested_layer=0):
top_layer = layers[0].get_weights()[0]
bot_layer = layers[requested_layer].get_weights()[0]
input_dim = top_layer.shape[0]
bot_dim = bot_layer.shape[1]
y = np.zeros([input_dim, bot_dim])
W = top_layer
for i in range(0, input_dim):
w_norm = np.sqrt(sum(np.square(W[i, :])))
y[i, :] = W[i, :] / w_norm
return y, input_dim, bot_dim
def plotTSNE(toPlot, labels, nb_classes, title = ""):
x_min, x_max = np.min(toPlot, 0), np.max(toPlot, 0)
toPlot = (toPlot - x_min) / (x_max - x_min)
print(toPlot.shape)
plt.figure()
y_unique = np.unique(labels)
cmap = sns.color_palette("Set2", n_colors=len(y_unique))
for i in range(0, len(y_unique)):
plt.scatter(toPlot[labels.as_matrix()==y_unique[i], 0], toPlot[labels.as_matrix()==y_unique[i], 1], c=cmap[i], label=y_unique[i], s=50)
plt.legend()
plt.xticks(())
plt.yticks(())
plt.tight_layout()
sns.despine(left=True, bottom=True)
plt.savefig("plots/" + datetime.now().strftime("%Y%m%dT%H%M%S") + "_" + title + '.png', bbox_inches='tight')
def plot3DScatter(toPlot, labels, nb_classes, title = ""):
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
x_min, x_max = np.min(toPlot, 0), np.max(toPlot, 0)
toPlot = (toPlot - x_min) / (x_max - x_min)
print(toPlot.shape)
y_unique = np.unique(labels)
cmap = sns.color_palette("Set2", n_colors=len(y_unique))
for i in range(0, len(y_unique)):
ax.scatter(toPlot[labels.as_matrix()==y_unique[i], 0], toPlot[labels.as_matrix()==y_unique[i], 1], toPlot[labels.as_matrix()==y_unique[i], 2], c=cmap[i], label=y_unique[i], s=50)
plt.legend()
plt.xticks(())
plt.yticks(())
plt.tight_layout()
# sns.despine(left=True, bottom=True)
plt.savefig("plots/" + datetime.now().strftime("%Y%m%dT%H%M%S") + "_" + title + '.png', bbox_inches='tight')
RANDOM_SEED = 859438905
def one_hot_dataframe(data, cols, replace=False):
vecData = pd.get_dummies(data, columns=cols)
return vecData
def preprocess_data(X, scaler=None):
if not scaler:
scaler = StandardScaler()
scaler.fit(X)
X = scaler.transform(X)
return X, scaler
X = getDefaultData("10000")
X = X.append(getNonDefaultData(str(len(X))))
X = X.drop(["channel", "loan_purpose", "property_type", "occupancy_status", "original_loan_term", "number_of_units", "prepayment_penalty_flag", "first_time_homebuyer_flag"], axis=1)
# X = X[["credit_score", "hpi_at_origination", "dti", "default_flag"]]
print("Data is equally matched over defaults and non-defaults. Size of data: " + str(len(X)))
# X = one_hot_dataframe(X, ["channel", "loan_purpose", "property_type", "occupancy_status"], replace=True)
# Shuffle
idx = np.random.permutation(len(X))
X = X.iloc[idx]
# Target
y = X["default_flag"]
del X["default_flag"]
## END OF PANDAS DF
# Hot one encode
X = one_hot_dataframe(X, ["number_of_borrowers"])
# Replace NaNs with imputed values
imp = Imputer(missing_values=0, strategy='mean', axis=0, copy=False)
X_imp = imp.fit_transform(X)
# Then scale
X_norm, scaler = preprocess_data(X_imp)
# Then put imputed values back to zero
X_norm[X.as_matrix()==0] = 0
# Split the dataset in two equal parts
X_train, X_test, Y_train, y_test = train_test_split(
X_norm, y, test_size=0.1, random_state=0)
nb_classes = 2
Y_train = np_utils.to_categorical(Y_train, nb_classes)
Y_test = np_utils.to_categorical(y_test, nb_classes)
batch_size = 128
activation = 'sigmoid'
input_dim = X_train.shape[1]
hidden_dim = 10
final_dim = 3
nb_epoch = 500
print("X_train: ", X_train.shape)
print("X_test: ", X_test.shape)
print("Y_train: ", Y_train.shape)
print("Y_test: ", Y_test.shape)
# print("Performing TSNE")
# model = TSNE(n_components=2, random_state=0, init="pca")
# toPlot = model.fit_transform(X_test)
# plotTSNE(toPlot, y_test, nb_classes, "1_t-SNE embedding for AutoEncoded output")
# Visualize result using PCA
print("Performing PCA")
pca = TruncatedSVD(n_components=2)
X_reduced = pca.fit_transform(X_test)
plotTSNE(X_reduced, y_test, nb_classes, "2_PCA for AutoEncoded output")
##########################
# dense model test #
##########################
# print("Training classical fully connected layer for classification")
# model_classical = Sequential()
# model_classical.add(Dense(input_dim, 2, activation=activation))
# model_classical.add(Dropout(0.3))
# model_classical.add(Activation('softmax'))
# model_classical.compile(loss='binary_crossentropy', optimizer='adadelta')
# model_classical.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=0, validation_data=(X_test, Y_test))
# classical_score = model_classical.evaluate(X_test, Y_test, verbose=0, show_accuracy=True)
# print('classical_score:', classical_score)
#
# y_true, y_pred = y_test, model_classical.predict_classes(X_test)
# print(classification_report(y_true, y_pred))
##########################
# autoencoder model test #
##########################
# plotExamples(X_train, number = 25, title="Original_")
# print("Training AutoEncoder for feature viz")
# AutoEncoder for feature visualization
encoder = containers.Sequential([Dense(input_dim, hidden_dim, activation=activation), Dense(hidden_dim, final_dim, activation=activation)])
decoder = containers.Sequential([Dense(final_dim, hidden_dim, activation=activation), Dense(hidden_dim, input_dim, activation=activation)])
# autoencoder.add(AutoEncoder(encoder=encoder, decoder=decoder, output_reconstruction=True))
# autoencoder.compile(loss='mse', optimizer='adadelta')
# # Do NOT use validation data with return output_reconstruction=True
# autoencoder.fit(X_train, X_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=0)
# # plotExamples(autoencoder.predict(X_train), number = 25, title="Reproduction_")
# vals = getActivationsLayer0(autoencoder.layers[0].encoder.layers)[0].T
# # plotExamples(vals, number = hidden_dim, title="Neuron_features_")
print("Training AutoEncoder for TSNE and classification")
# AutoEncoder for TSNE and classification
autoencoder = Sequential()
autoencoder.add(AutoEncoder(encoder=encoder, decoder=decoder, output_reconstruction=False))
autoencoder.compile(loss='mean_squared_error', optimizer='adam')
# Do NOT use validation data with return output_reconstruction=True
autoencoder.fit(X_train, X_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=1)
# Do an inference pass
prefilter_train = autoencoder.predict(X_train, verbose=0)
prefilter_test = autoencoder.predict(X_test, verbose=0)
print("prefilter_train: ", prefilter_train.shape)
print("prefilter_test: ", prefilter_test.shape)
print("Performing TSNE")
model = TSNE(n_components=2, random_state=0, init="pca")
toPlot = model.fit_transform(prefilter_test)
plotTSNE(toPlot, y_test, nb_classes, "3_t-SNE embedding for AutoEncoded output")
plot3DScatter(prefilter_test, y_test, nb_classes, "5_3D_Raw_AutoEncoded output")
# Visualize result using PCA
pca = TruncatedSVD(n_components=2)
X_reduced = pca.fit_transform(prefilter_test)
plotTSNE(X_reduced, y_test, nb_classes, "4_PCA for AutoEncoded output")
#
# print("Classifying and comparing")
# # Classify results from Autoencoder
# model = Sequential()
# model.add(Dense(prefilter_train.shape[1], nb_classes, activation=activation))
#
# model.add(Activation('softmax'))
#
# model.compile(loss='binary_crossentropy', optimizer='adadelta')
# model.fit(prefilter_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch, show_accuracy=False, verbose=0, validation_data=(prefilter_test, Y_test))
#
# score = model.evaluate(prefilter_test, Y_test, verbose=0, show_accuracy=True)
# print('\nscore:', score)
#
# print('Loss change:', 100*(score[0] - classical_score[0])/classical_score[0], '%')
# print('Accuracy change:', 100*(score[1] - classical_score[1])/classical_score[1], '%')
#
# y_true, y_pred = y_test, model.predict_classes(prefilter_test)
# print(classification_report(y_true, y_pred))