-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmake_figs_classification.py
246 lines (221 loc) · 9.49 KB
/
make_figs_classification.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
# -*- coding: utf-8 -*-
#
# Copyright 2020 Pietro Barbiero, Alberto Tonda and Giovanni Squillero
# Licensed under the EUPL
import copy
import glob
import json
import os
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import matplotlib.patches as mp
from lazygrid.datasets import load_openml_dataset
from matplotlib.colors import ListedColormap
from sklearn.datasets import make_classification
from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
from sklearn.feature_selection import RFE
from sklearn.linear_model import RidgeClassifier
from sklearn.metrics import f1_score
from sklearn.model_selection import cross_validate, StratifiedKFold
from sklearn.preprocessing import StandardScaler
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
random_state = 42
n_jobs = 2
cv = 2
results_dir = "./results/classification/"
file_list = glob.glob(results_dir + '**verbose.csv')
test_score = []
train_score = []
feature_size = []
feature_selector = []
estimator = []
fit_time = []
dataset_name = []
for fi, file in enumerate(file_list):
# if ('yeast' not in file) and ('scene' not in file):
# continue
dataset = file.split('/')[3].split('_verbose.csv')[0]
X, y, n_classes = load_openml_dataset(dataset_name=dataset)
# X, y = datasets[dataset]
results = pd.read_csv(file, index_col=0)
features_selected = results['features_selected']
n_selected = []
for r in range(results.shape[0]):
print(f'{fi+1}/{len(file_list)} - {r+1}/{results.shape[0]}')
selector = results['estimator'][r]
time = results['fit_time'][r]
feature_set_str = features_selected[r].replace('[', '').replace(']', '').split()
feature_set = np.array([int(i) for i in feature_set_str])
skf = StratifiedKFold(n_splits=10, shuffle=True, random_state=42)
train_index, test_index = [split for split in skf.split(X, y)][r % 10]
sc = StandardScaler()
sc.fit(X[train_index])
X_train = sc.transform(X[train_index])
X_test = sc.transform(X[test_index])
if selector == 'DFE':
X_train = X_train[:, feature_set == 1]
X_test = X_test[:, feature_set == 1]
n_selected.append(X_train.shape[1])
feature_set_bool = feature_set
elif selector == 'RFE':
X_train = X_train[:, feature_set == 1]
X_test = X_test[:, feature_set == 1]
n_selected.append(X_train.shape[1])
feature_set_bool = feature_set
# rfe = RFE(RidgeClassifier(random_state=random_state), n_features_to_select=n_selected[r % 10])
# rfe.fit(X_train, y[train_index])
# X_train = X_train[:, rfe.ranking_ == 1]
# X_test = X_test[:, rfe.ranking_ == 1]
elif selector == 'NO-FS':
X_train = X_train
X_test = X_test
else:
X_train = X_train[:, feature_set[:n_selected[r % 10]]]
X_test = X_test[:, feature_set[:n_selected[r % 10]]]
feature_set_sorted = np.argsort(feature_set)
feature_set_bool = np.zeros((feature_set.shape[0],)).astype(int)
feature_set_bool[:n_selected[r % 10]] = 1
feature_set_bool = feature_set_bool[feature_set_sorted]
classifiers = [
# RandomForestClassifier(random_state=random_state),
RidgeClassifier(random_state=random_state),
DecisionTreeClassifier(random_state=random_state),
# SVC(random_state=random_state),
]
if selector == 'DFE':
selector = 'PFE'
for classifier in classifiers:
classifier.fit(X_train, y[train_index])
train_pred = classifier.predict(X_train)
test_pred = classifier.predict(X_test)
train_score_r = f1_score(y[train_index], train_pred, average='weighted')
test_score_r = f1_score(y[test_index], test_pred, average='weighted')
test_score.append(test_score_r)
train_score.append(train_score_r)
feature_size.append(X_train.shape[1])
feature_selector.append(selector)
estimator.append(classifier.__class__.__name__)
fit_time.append(time)
dataset_name.append(dataset)
results = pd.DataFrame({
'dataset': dataset_name,
'test F1': test_score,
'train_score': train_score,
'feature_size': feature_size,
'algorithm': feature_selector,
'estimator': estimator,
'training time (s)': fit_time,
})
import matplotlib
sns.set_style('whitegrid')
for dataset in set(results['dataset']):
# miny = results[results['dataset'] == dataset]['test F1'].min() - 0.005
# maxy = results[results['dataset'] == dataset]['test F1'].max() + 0.005
for estimator in set(results['estimator']):
res = results[(results['dataset'] == dataset) & (results['estimator'] == estimator)]
# res = res.sort_values(['algorithm'])
perc_test_f1 = []
for algorithm in sorted(set(res['algorithm'])):
perc_test_f1.extend(res[res['algorithm'] == algorithm]['test F1'].values - res[res['algorithm'] == 'NO-FS']['test F1'].values)
res['% test F1'] = perc_test_f1
resPFE = res[res['algorithm']=='PFE']
resRFE = res[res['algorithm']=='RFE']
resALL = res[(res['algorithm']!='PFE') & (res['algorithm']!='RFE') & (res['algorithm']!='NO-FS')]
res = pd.concat([resPFE, resALL, resRFE])
plt.figure(figsize=[4, 3])
sns.scatterplot(x='training time (s)', y='% test F1', hue='algorithm', data=res, alpha=0.4)
sns.scatterplot(x='training time (s)', y='% test F1', hue='algorithm', data=res[res['algorithm']=='PFE'], alpha=1)
plt.legend().set_visible(False)
plt.title(f'{dataset} - {estimator}')
# plt.ylim([miny, maxy])
plt.xscale('log')
plt.tight_layout()
plt.savefig(f'{results_dir}{dataset}_{estimator}_classification.png')
plt.savefig(f'{results_dir}{dataset}_{estimator}_classification.pdf')
plt.show()
plt.figure(figsize=[4, 5])
sns.scatterplot(x='training time (s)', y='test F1', hue='algorithm', data=res, alpha=0.4)
plt.legend(loc='upper center', bbox_to_anchor=(0.5, -0.8), ncol=3)
plt.tight_layout()
plt.savefig(f'{results_dir}legend_classification.png', bbox_inches=matplotlib.transforms.Bbox([[0.4, 2.4], [4, 3.3]]))
plt.savefig(f'{results_dir}legend_classification.pdf', bbox_inches=matplotlib.transforms.Bbox([[0.4, 2.4], [4, 3.3]]))
plt.show()
# results = pd.read_csv(results_file, index_col=0)
# results["avg test error"] = 1 - results["avg test score"]
# results["sem test error"] = results["sem test score"]
# results["size"] = results["samples"] * results["features"]
# dataset_set = results[["dataset", "samples", "features", "size"]].drop_duplicates().sort_values("size")
# n_rows = int(np.sqrt(len(dataset_set)))
# dataset_set.drop([0, 48, 64], axis=0, inplace=True)
#
# res = results.copy()
# ok_list = [
# "yeast_ml8",
# "scene",
# "isolet",
# "gina_agnostic",
# "gas-drift",
# "letter",
# ]
# idx = []
# for i in range(len(res)):
# if res["dataset"][i] in ok_list:
# idx.append(i)
# res = res.iloc[idx]
# res["size"] = res["samples"] * res["features"]
#
# plt.figure()
# sns.lineplot(x="size", y="avg fit time", data=res, hue="method")
# plt.xscale("log")
# plt.yscale("log")
# plt.legend(loc='upper center', bbox_to_anchor=(0.5, -0.3), ncol=3)
# plt.show()
# def is_pareto_efficient_simple(costs):
# """
# Find the pareto-efficient points
# :param costs: An (n_points, n_costs) array
# :return: A (n_points, ) boolean array, indicating whether each point is Pareto efficient
# """
# is_efficient = np.ones(costs.shape[0], dtype = bool)
# for i, c in enumerate(costs):
# if is_efficient[i]:
# is_efficient[is_efficient] = np.any(costs[is_efficient]<c, axis=1) # Keep any point with a lower cost
# is_efficient[i] = True # And keep self
# return is_efficient
#
#
# plt.figure(figsize=[8, 12])
# for i, (k, dataset) in enumerate(dataset_set.iterrows()):
# d = results[results["dataset"] == dataset["dataset"]]
# is_pareto = is_pareto_efficient_simple(d[d["method"] != "NO-FS"][["avg fit time", "avg test error"]].values)
# current_palette = sns.color_palette(n_colors=d.shape[0])
# plt.subplot(3, 2, i+1)
# lines = []
# labels = []
# for xi, yi, exi, eyi, method, c in zip(d["avg fit time"], d["avg test error"], d["sem fit time"], d["sem test error"], d["method"], current_palette):
# l = plt.errorbar(xi, yi, xerr=exi, yerr=eyi, fmt='o', ecolor=c, c=c, alpha=0.5)
# e = mp.Ellipse((xi, yi), width=exi*2, height=eyi*2, color=c, alpha=0.5)
# plt.gca().add_patch(e)
# lines.append(l)
# labels.append(method)
# pareto_x = d[d["method"] != "NO-FS"]["avg fit time"][is_pareto].values
# pareto_y = d[d["method"] != "NO-FS"]["avg test error"][is_pareto].values
# x_sorted = np.argsort(pareto_x)
# pareto_x = pareto_x[x_sorted]
# pareto_y = pareto_y[x_sorted]
# plt.plot(pareto_x, pareto_y, c='k', ls='--', alpha=0.8)
# plt.scatter(pareto_x, pareto_y, c='k', marker='*', s=80)
# plt.title(f'{dataset["dataset"]} [S={dataset["samples"]}, F={dataset["features"]}]')
# plt.xscale("log")
# if i >= 4:
# plt.xlabel("fit time")
# if i % 2 == 0:
# plt.ylabel("test error")
# plt.legend(lines, labels, loc='upper center', bbox_to_anchor=(0.5, -0.3), ncol=3)
# plt.tight_layout()
# plt.savefig("summary.png")
# plt.savefig("summary.pdf")
# plt.show()