/
dataset1.py
257 lines (201 loc) · 9.16 KB
/
dataset1.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
253
254
255
256
257
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.spatial import distance
from scipy.stats import kurtosis
import os
from time import clock
# ml package
from sklearn.cross_validation import cross_val_score
from sklearn.cross_validation import train_test_split
from pylab import rcParams
from DataLoader import DataLoader
from sklearn.cluster import KMeans
from sklearn.mixture import GaussianMixture
from sklearn.decomposition import PCA, FastICA
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.random_projection import SparseRandomProjection
from sklearn.neural_network import MLPClassifier
rcParams['figure.figsize'] = 10, 7
# load data
output_path = 'outputs\\Marketing'
dl_1 = DataLoader('data\\UCI-bank-marketing.csv', output_path, 'Marketing')
dl_1.load_data()
dl_1.scaled_data()
X, y = dl_1.get_data()
# k-means clustering
# Expectation Maximization
# PCA
# ICA
# Randomized Projections
# LDA
def clustering_algo(X, y, cluster, n_c=2, n_i=10):
if cluster == 'KM':
clf = KMeans(n_clusters=n_c, n_init=n_i).fit(X)
elif cluster == 'EM':
clf = GaussianMixture(n_components=n_c, n_init=n_i).fit(X)
y_pred = clf.predict(X)
return (max(sum(y == y_pred) / len(y), 1 - sum(y == y_pred) / len(y)))
def dim_reduce(X, y, algo):
if algo == 'PCA':
transformer = PCA(n_components=2)
elif algo == 'ICA':
transformer = FastICA(n_components=2)
elif algo == 'SparseRandomProjection':
transformer = SparseRandomProjection(n_components=2)
elif algo == 'LDA':
transformer = LinearDiscriminantAnalysis(n_components=2)
return transformer.fit(X, y).transform(X), transformer.fit(X, y)
# part 1
acc_km = clustering_algo(X, y, cluster='KM', n_c=2, n_i=10)
acc_em = clustering_algo(X, y, cluster='EM', n_c=2, n_i=30)
print('benchmark', max(sum(y) / len(y), 1 - sum(y) / len(y)))
print('acc_km: ', acc_km)
print('acc_em: ', acc_em)
# part 2
x_PCA, n_pca = dim_reduce(X, y, algo='PCA')
x_ICA, n_ica = dim_reduce(X, y, algo='ICA')
x_SRP, n_srp = dim_reduce(X, y, algo='SparseRandomProjection')
x_LDA, n_lda = dim_reduce(X, y, algo='LDA')
print('PCA Eigenvalues', n_pca.explained_variance_ratio_)
print('ICA Kurtosis', kurtosis(x_ICA))
# part 3
df_acc_label, df_acc_comp, df_acc_km, df_acc_em = [], [], [], []
for name, X_temp in zip(['Raw', 'PCA', 'ICA', 'SRP', 'LDA'], [X, x_PCA, x_ICA, x_SRP, x_LDA]):
acc_km = clustering_algo(X_temp, y, cluster='KM', n_c=2, n_i=10)
acc_em = clustering_algo(X_temp, y, cluster='EM', n_c=2, n_i=30)
df_acc_label.append(name)
df_acc_comp.append(len(X.T))
df_acc_km.append(acc_km)
df_acc_em.append(acc_em)
print(name + ': components ' + str(len(X.T)))
# print('benchmark', max(sum(y)/len(y), 1-sum(y)/len(y)))
print('acc_km: ', acc_km)
print('acc_em: ', acc_em)
df_acc = pd.DataFrame(data={'Algo': df_acc_label, '# of Component': df_acc_comp, 'KM Accuracy': df_acc_km, 'EM Accuracy': df_acc_em}, columns=['Algo', '# of Component', 'KM Accuracy', 'EM Accuracy'])
def compute_aic_bic(kmeans, X):
"""
Computes the BIC metric for a given clusters
Parameters:
-----------------------------------------
kmeans: List of clustering object from scikit learn
X : multidimension np array of data points
Returns:
-----------------------------------------
BIC value
"""
# assign centers and labels
centers = [kmeans.cluster_centers_]
labels = kmeans.labels_
# number of clusters
m = kmeans.n_clusters
# size of the clusters
n = np.bincount(labels)
# size of data set
N, d = X.shape
# compute variance for all clusters beforehand
cl_var = (1.0 / (N - m) / d) * sum([sum(distance.cdist(X[np.where(labels == i)], [centers[0][i]],
'euclidean') ** 2) for i in range(m)])
ln_likelihood = np.sum([n[i] * np.log(n[i]) -
n[i] * np.log(N) -
((n[i] * d) / 2) * np.log(2 * np.pi * cl_var) -
((n[i] - 1) * d / 2) for i in range(m)])
AIC = 2 * m - 2 * ln_likelihood
BIC = m * np.log(N) * (d + 1) - 2 * ln_likelihood
return AIC, BIC
X = np.array(X)
def best_km_cluster(X, max_cluster=None, title=None):
ks = range(1, max_cluster)
kms = [KMeans(n_clusters=i, init="k-means++").fit(X) for i in ks]
clst, aic, bic = [], [], []
for i in range(len(kms)):
temp = compute_aic_bic(kms[i], X)
clst.append(ks[i])
aic.append(temp[0])
bic.append(temp[1])
df_cluster = pd.DataFrame(data={'cluster': clst, 'aic': aic, 'bic': bic}, columns=['cluster', 'aic', 'bic'])
plt.close()
plt.figure()
plt.plot(df_cluster['cluster'], df_cluster['aic'], '-o', label='AIC')
plt.plot(df_cluster['cluster'], df_cluster['bic'], '-o', label='BIC')
plt.grid()
plt.legend()
if title == None:
plt.title('K Means')
else:
plt.title('K Means:' + title)
plt.savefig(os.path.join(output_path, '{}_best_KM.png'.format(title)), dpi=150)
return df_cluster
def best_em_cluster(X, max_cluster=None, title=None):
ks = range(1, max_cluster)
gmm = [GaussianMixture(n_components=i).fit(X) for i in ks]
clst, aic, bic = [], [], []
for i in range(len(gmm)):
# temp = compute_aic_bic(KMeans[i],X)
clst.append(ks[i])
aic.append(gmm[0].aic(X))
bic.append(gmm[0].bic(X))
df_cluster = pd.DataFrame(data={'cluster': clst, 'aic': aic, 'bic': bic}, columns=['cluster', 'aic', 'bic'])
plt.close()
plt.figure()
plt.plot(df_cluster['cluster'], df_cluster['aic'], '-o', label='AIC')
plt.plot(df_cluster['cluster'], df_cluster['bic'], '-o', label='BIC')
plt.grid()
plt.legend()
if title == None:
plt.title('EM')
else:
plt.title('EM:' + title)
plt.savefig(os.path.join(output_path, '{}_best_EM.png'.format(title)), dpi=150)
return df_cluster
df_cluster_km = best_km_cluster(X, max_cluster=10, title=None)
df_cluster_em = best_em_cluster(X, max_cluster=10, title=None)
for name, X_temp in zip(['Raw', 'PCA', 'ICA', 'SRP', 'LDA'], [X, x_PCA, x_ICA, x_SRP, x_LDA]):
best_km_cluster(np.array(X_temp), max_cluster=10, title=name)
best_em_cluster(np.array(X_temp), max_cluster=10, title=name)
# part 4 and 5
def NN(training_data, training_labels, test_data, test_labels, cv=True, hls=(100, 2), mxit=500, act='relu', isr=False):
if cv:
depth = pd.DataFrame(columns=['Layer', 'Iteration', 'cv_scores', 'Pred_Acc'])
for j in range(4, 15):
for i in range(1, 30):
clf_nn = MLPClassifier(hidden_layer_sizes=(i * 10, 2), activation=act, solver='adam', max_iter=j * 50)
# Perform 7-fold cross validation
scores = cross_val_score(estimator=clf_nn, X=training_data, y=training_labels, cv=10, n_jobs=4)
# pred
clf_nn.fit(training_data, training_labels)
pred = np.array(clf_nn.predict(test_data))
success_rate = len(test_labels[test_labels == pred]) / len(test_labels)
depth = depth.append([{'Layer': i * 10, 'Iteration': j * 50, 'cv_scores': scores.mean(), 'Pred_Acc': success_rate}], ignore_index=True)
cv_best = depth.sort_values(by='cv_scores', ascending=False).iloc[0, :]
pred_best = depth.sort_values(by='Pred_Acc', ascending=False).iloc[0, :]
print('CV best: with layer' + str(2) + ' \n', cv_best)
print('Pred best: with layer' + str(2) + ' \n', pred_best)
return depth, cv_best, pred_best
else:
clf_nn = MLPClassifier(hidden_layer_sizes=hls, activation=act, solver='adam', max_iter=mxit)
clf_nn.fit(training_data, training_labels)
pred = np.array(clf_nn.predict(test_data))
success_rate = len(test_labels[test_labels == pred]) / len(test_labels)
print("NN: ", success_rate, ', with hidden_layer_sizes=', hls, ', with max_iter=', mxit, ' and activation=', act)
if isr:
pred_train = np.array(clf_nn.predict(training_data))
srt = len(training_labels[training_labels == pred_train]) / len(training_labels)
return srt, success_rate
else:
return success_rate
X_KM = KMeans(n_clusters=2, n_init=10).fit(X).predict(X)
X_EM = GaussianMixture(n_components=2, n_init=10).fit(X).predict(X)
X_KM = np.append(X, X_KM.reshape((-1, 1)), axis=1)
X_EM = np.append(X, X_EM.reshape((-1, 1)), axis=1)
df_sr_label, df_sr_train, df_sr_test, df_sr_time = [], [], [], []
for name, x_temp in zip(['Raw', 'PCA', 'ICA', 'SRP', 'LDA', 'KM', 'EM'], [X, x_PCA, x_ICA, x_SRP, x_LDA, X_KM, X_EM]):
training_data, test_data, training_labels, test_labels = train_test_split(x_temp, y, test_size=0.4, random_state=0)
st = clock()
sr_train, sr_test = NN(training_data, training_labels, test_data, test_labels, cv=False, hls=(26, 26, 26), mxit=300, act='relu', isr=True)
df_sr_time.append(clock() - st)
df_sr_label.append(name)
df_sr_train.append(sr_train)
df_sr_test.append(sr_train)
df_sr = pd.DataFrame(data={'Algo': df_sr_label, 'Testing Accuracy': df_sr_test, 'Time': df_sr_time}, columns=['Algo', 'Testing Accuracy', 'Time'])
print(df_acc, df_sr)