-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathMIAMI_distribution_analysis.py
277 lines (193 loc) · 10 KB
/
MIAMI_distribution_analysis.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
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
# -*- coding: utf-8 -*-
"""
Created on Fri Jun 4 15:15:52 2021
@author: rfuchs
"""
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.preprocessing import LabelEncoder
from dython.nominal import compute_associations
from table_evaluator import TableEvaluator
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import f1_score
os.chdir(r'C:\Users\rfuchs\Documents\These\Stats\mixed_dgmm\datasets')
train_test_folder = 'adult'
fake_folder = 'Adult data'
varnames = ['age', 'workclass', 'fnlwgt', 'education',
'education.num', 'marital.status', 'occupation', 'relationship',\
'race', 'sex', 'capital.gain', 'capital.loss',\
'hours.per.week', 'native.country', 'income']
var_distrib = np.array(['continuous', 'categorical', 'continuous', 'ordinal',\
'binomial', 'categorical', 'categorical', 'categorical',\
'categorical', 'bernoulli', 'ordinal', 'ordinal',\
'continuous', 'categorical', 'bernoulli'])
p = len(var_distrib)
cat_features = (pd.Series(var_distrib) != 'continuous').to_list()
dtype = {varnames[j]: np.str if cat_features[j]
else np.float64 for j in range(p)}
#====================================
# Make some 3D plots with matplotlib
#====================================
designs = ['Absent', 'Small', 'Unbalanced']
datasets = ['preds' + str(i) + '.csv' for i in range(1,11)]
x = 'capital.loss'
y = 'fnlwgt'
z = 'hours.per.week'
f1_scores_dict = {}
for design in designs:
models = os.listdir(fake_folder + '/' + design)
nb_models = len(models)
data_prefix = '' if design == 'Small' else design[:3] + '_'
nb_plots = 3 if design != 'Absent' else 2
offset = 1 if design != 'Absent' else 0 # To take into account the lack of train data for Absent
f1_scores = pd.DataFrame(columns = ['set_idx', 'classifier', 'model', 'f1'])
for idx in range(1,11):
train = pd.read_csv(train_test_folder + '/' + design + '/' +\
'/' + data_prefix + 'Train' + str(idx) + '.csv')
test = pd.read_csv(train_test_folder + '/' + design + '/' +\
'/' + data_prefix + 'Test_' + str(idx) + '.csv')
#==========================================
# Trivariate distributions
#==========================================
# Keep only the interesting modalities
if design in ['Absent', 'Unbalanced']:
train = train[(train['age'] >= 60) & (train['sex'] == 'Female')]
test = test[(test['age'] >= 60) & (test['sex'] == 'Female')]
data = {}
for model_idx, model in enumerate(models):
try:
fake = pd.read_csv(fake_folder + '/' + design + '/' + model +\
'/preds' + str(idx) + '.csv')
data[model] = fake
except FileNotFoundError:
print(model, ' has no preds ', str(idx))
fake = pd.DataFrame(columns = train.columns)
data['Train'] = train
data['Test'] = test
plot_datasets = list(set(data.keys()))
plot_datasets.sort()
plot_datasets = [d for d in plot_datasets\
if (d != 'Train') & (d != 'Test')] + ['Test']
if design != 'Absent':
plot_datasets = ['Train'] + plot_datasets
fig = plt.figure(figsize = plt.figaspect(1 / len(plot_datasets)))
#fig.suptitle(design + ' ' + ' (preds' + str(idx) + ')',\
#fontsize=16)
for i, dataset in enumerate(plot_datasets):
if (design == 'Absent') & (dataset == "Train"):
continue
ax = fig.add_subplot(1, len(plot_datasets), i + 1, projection='3d')
ax.scatter(data[dataset][x], data[dataset][y], data[dataset][z])
ax.set_xlabel(x)
ax.set_ylabel(y)
ax.set_zlabel(z)
ax.set_xlim(data['Test'][x].min() - 10,data['Test'][x].max() + 10)
ax.set_ylim(data['Test'][y].min() - 10,data['Test'][y].max() + 10)
ax.set_zlim(data['Test'][z].min() - 10,data['Test'][z].max() + 10)
ax.set_title(dataset, fontsize = 24)
#fig.savefig('../Figures/adult/' + design + '/preds' + str(idx) + '.png')
plt.show()
# Dirty hack to remove
if design == 'Unbalanced':
fig = plt.figure(figsize = plt.figaspect(1 / 30))
for i, dataset in enumerate(plot_datasets[:5]):
ax = fig.add_subplot(1, len(plot_datasets), i + 1, projection='3d')
ax.scatter(data[dataset][x], data[dataset][y], data[dataset][z])
ax.set_xlabel(x)
ax.set_ylabel(y)
ax.set_zlabel(z)
ax.set_xlim(data['Test'][x].min() - 10,data['Test'][x].max() + 10)
ax.set_ylim(data['Test'][y].min() - 10,data['Test'][y].max() + 10)
ax.set_zlim(data['Test'][z].min() - 10,data['Test'][z].max() + 10)
if i == 4:
ax.set_xlabel(x)
ax.set_ylabel(y)
ax.set_zlabel(z)
ax.set_title(dataset, fontsize = 14)
plt.tight_layout()
plt.show()
fig = plt.figure(figsize = plt.figaspect(1 / 5))
for i, dataset in enumerate(plot_datasets[5:]):
ax = fig.add_subplot(1, len(plot_datasets), i + 1, projection='3d')
ax.scatter(data[dataset][x], data[dataset][y], data[dataset][z])
ax.set_xlim(data['Test'][x].min() - 10,data['Test'][x].max() + 10)
ax.set_ylim(data['Test'][y].min() - 10,data['Test'][y].max() + 10)
ax.set_zlim(data['Test'][z].min() - 10,data['Test'][z].max() + 10)
if i == 4:
ax.set_xlabel(x)
ax.set_ylabel(y)
ax.set_zlabel(z)
ax.set_title(dataset, fontsize = 14)
plt.tight_layout()
plt.show()
dataset = 'MIAMI'
cols= ['age', 'workclass', 'marital.status']
table_evaluator = TableEvaluator(data['Test'][cols], data[dataset][cols])
table_evaluator.plot_distributions()
#==========================================
# Correlation matrices
#==========================================
w_ratios = [1] * len(plot_datasets) + [0.08]
fig, axs = plt.subplots(1, len(plot_datasets) + 1, \
gridspec_kw={'width_ratios':w_ratios}, figsize = (15,10))
axs[0].get_shared_y_axes().join(*axs[1:-1])
#fig.suptitle(design + ' ' + ' (preds' + str(idx) + ')',\
#fontsize=16)
for i, dataset in enumerate(plot_datasets):
if (design == 'Absent') & (dataset == "Train"):
continue
# Delete useless variables
data[dataset] = data[dataset].iloc[:, :p] # Delete legacy columns
# Encode categorical data
for col_idx, colname in enumerate(train.columns):
if cat_features[col_idx]:
le = LabelEncoder()
data[dataset][colname] = le.fit_transform(data[dataset][colname])
# Delete the sex column
data[dataset] = data[dataset][[name for name in varnames if name != 'sex']]
cat_features_new = [cat_features[i] for i in range(len(varnames)) if varnames[i] != 'sex']
if i == len(plot_datasets) - 1:
corr = compute_associations(data[dataset].astype(int), nominal_columns = cat_features_new)
g1 = sns.heatmap(corr,cmap="YlGnBu",ax=axs[i], cbar_ax=axs[-1])
else:
corr = compute_associations(data[dataset].astype(int), nominal_columns = cat_features_new)
g1 = sns.heatmap(corr,cmap="YlGnBu",cbar=False,ax=axs[i])
g1.set_ylabel('')
g1.set_xlabel('')
g1.set_yticks([])
axs[i].set_title(dataset, fontsize = 25)
plt.tight_layout()
plt.show()
#==========================================
# Predictive performances
#==========================================
target = 'education.num'
covariates = data['Test'].columns.tolist()
covariates.remove(target)
metric = 'weighted'
X_test = data['Test'][covariates]
y_test = data['Test'][target]
for i, dataset in enumerate(plot_datasets[:-1]):
if (design == 'Absent') & (dataset == "Train"):
continue
X_train = data[dataset].round(0).astype(int)[covariates]
y_train = data[dataset].round(0).astype(int)[target]
clf = RandomForestClassifier(max_depth=2, random_state=0)
clf.fit(X_train, y_train)
pred = clf.predict(X_test)
print(np.unique(pred, return_counts=True))
f1 = f1_score(y_test, pred, average=metric)
f1_scores = f1_scores.append({'set_idx': idx, 'classifier': 'RF',\
'model': dataset, 'f1': f1}, ignore_index = True)
f1_scores_dict[design] = f1_scores
for key, item in f1_scores_dict.items():
ax = sns.boxplot(x = "model", y = "f1",
data=item, palette="Set3",\
showfliers = False)
ax.tick_params('x', rotation = 30)
ax.set_ylabel(metric + 'precision')
ax.set_title('[' + key + ']' + metric + ' precision to predict ' + target + 'with a RF classifier')
plt.show()