-
Notifications
You must be signed in to change notification settings - Fork 2
/
msrigs_drawfig4_sensitivityanalysis.py
252 lines (235 loc) · 11.9 KB
/
msrigs_drawfig4_sensitivityanalysis.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
# Multi-Stage Re-Identification (MSRI) Game Solver (GS) v1.1
# Component: Showing sensitivity analysis results in line plots (payoff, privacy, and utility) and violin plots (payoff)
# © Oct 2018-2021 Zhiyu Wan, HIPLAB, Vanderbilt University
# Compatible with python 3.8.5. Package dependencies: Numpy 1.19.1, Pandas 1.1.3, Matplotlib 3.3.1, Seaborn 0.11.0
# Update history:
# April 21, 2020: plot multiple figure
# Aug 1, 2020: plot all sub-figures
# Oct 13, 2020: Add two variations of masking game
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import time
start1 = time.time()
SMALL_SIZE = 8
MEDIUM_SIZE = 10
BIGGER_SIZE = 12
plt.rc('font', size=SMALL_SIZE) # controls default text sizes
plt.rc('axes', titlesize=SMALL_SIZE) # fontsize of the axes title
plt.rc('axes', labelsize=9) # fontsize of the x and y labels (original: MEDIUM_SIZE)
plt.rc('xtick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('ytick', labelsize=SMALL_SIZE) # fontsize of the tick labels
plt.rc('legend', fontsize=SMALL_SIZE) # legend fontsize
plt.rc('figure', titlesize=BIGGER_SIZE) # fontsize of the figure title
id_exp = '2056' # ID for the set of experiments
n_iter = 20
n_S = 1000
n_scenario = 8
pruning = 1
order = [0, 1, 2, 3, 7, 4, 5, 6] # 0: no protection. 1: no genomic data sharing. 2: random opt-in. 3: random masking.
# 4: opt-in game. 5: masking game. 6: no-attack masking game. 7: one-stage masking game.
n_fig = 8 # Number of lineplot figures
n_row = [4, 3, 3] # number of rows of subplots in each figure
n_col = [3, 3, 3] # number of collums of subplots in each figure
fig_row = np.zeros(3).astype(int)
fig_col = np.zeros(3).astype(int)
fig0, axes0 = plt.subplots(n_row[0], n_col[0], figsize=(10.28, 12.6))
fig1, axes1 = plt.subplots(n_row[1], n_col[1], figsize=(10.4, 9.6))
fig2, axes2 = plt.subplots(n_row[2], n_col[2], figsize=(10.4, 9.6))
fig = [fig0, fig1, fig2]
axes = [axes0, axes1, axes2]
metric_name = ['payoff', 'privacy', 'utility']
column_names = ['defender_optimal', 'privacy', 'utility']
scenario_name = ['No-protection', 'Demographics-only', 'Random opt-in', 'Random masking', 'Opt-in game', 'Masking game',
'No-attack masking game', 'One-stage masking game']
scenario_name = np.array(scenario_name)
scenario_name_in_order = scenario_name[order]
experiments_name = ['m', 'missinglevel', 'theta', 'nG', 'nI', 'loss', 'benefit', 'cost']
experiments_name = [experiments_name[i] + 'changing' for i in range(len(experiments_name))]
xlabels = ['Number of genomic attributes', 'Proportion of missing genomic data', 'Threshold for confidence score',
'Number of records in the genetic genealogy dataset', 'Number of records in the identified dataset',
'Loss from being re-identified ($)', 'Maximal benefit of sharing all data ($)', 'Cost of attack ($)']
style_order = ['Masking game', 'One-stage masking game', 'No-attack masking game', 'Opt-in game', 'Random masking',
'Random opt-in', 'Demographics-only', 'No-protection']
if n_scenario == 7:
style_order = ['Masking game', 'One-stage masking game', 'No-protection', 'Opt-in game', 'Random masking',
'Random opt-in', 'Demographics-only']
ms_g_start = [2, 0, 0, 1, 1, 0, 0, 0]
n_ms = [15, 10, 11, 20, 20, 17, 17, 17]
ms_g_end = [17, 10, 11, 21, 21, 17, 17, 17]
steps = [1, 0.1, 0.1, 2000, 2000, 25, 25, 10]
default_xs = [12, 0.3, 0.5, 20000, 20000, 150, 100, 10]
for fig_i in range(n_fig): # each subplot
experiment_name = experiments_name[fig_i]
xlabel = xlabels[fig_i]
m_g_start = ms_g_start[fig_i]
m_g_end = ms_g_end[fig_i]
step = steps[fig_i]
default_x = default_xs[fig_i]
for i in range(3):
fig_row[i] = fig_i // n_col[i]
fig_col[i] = fig_i % n_col[i]
# plot the default vertical lines
for i in range(3):
axes[i][fig_row[i], fig_col[i]].axvline(x=default_x, label='Default value', color='0.5', linestyle='--')
folder_result = 'Results' + id_exp + '/' + experiment_name + '/'
if pruning == 1:
folder_result += 'pruning/'
folder_output = 'Results' + id_exp + '/'
# input result data (payoff, privacy, and utility)
n_m = m_g_end - m_g_start
payoffs = []
privacy = []
utility = []
for jj in range(n_scenario):
j = order[jj] # re-order for plotting
dataset = pd.read_pickle(folder_result + 'result_p' + str(fig_i) + '_s' + str(j) + '.pickle')
for i in range(3):
data = np.array(dataset[column_names[i]])
shaped_dataset = np.reshape(data, (n_iter * n_m, n_S))
av_dataset = np.mean(shaped_dataset, axis=1)
if i == 0:
payoffs.extend(av_dataset)
elif i == 1:
privacy.extend(av_dataset)
else:
utility.extend(av_dataset)
m_g = []
for i in range(n_scenario):
for j in range(m_g_start, m_g_end):
label = [j * step for k in range(n_iter)]
m_g.extend(label)
scenarios = []
for i in range(n_scenario):
label = [scenario_name_in_order[i] for j in range(n_iter * n_m)]
scenarios.extend(label)
# plot each figure
results = [payoffs, privacy, utility]
ylabels = ['Average payoff ($)', 'Privacy', 'Utility']
colors = ["tab:red", "tab:orange", "tab:olive", "tab:green", "tab:cyan", "tab:blue", "tab:purple", "tab:gray"]
colors = np.array(colors)
colors_in_order = colors[order]
customPalette = sns.set_palette(sns.color_palette(colors_in_order))
markers_base = ["X", "d", "^", "v", "o", "s", "P", "D"]
markers = dict(zip(scenario_name, markers_base))
for i in range(3):
dataset = pd.DataFrame({ylabels[i]: results[i],
'Scenario': scenarios,
xlabel: m_g,
})
sns.lineplot(data=dataset, x=xlabel, y=ylabels[i], hue='Scenario', markers=markers, style='Scenario',
palette=customPalette, style_order=style_order, ax=axes[i][fig_row[i], fig_col[i]], ci='sd')
if fig_i == 1:
axes[i][fig_row[i], fig_col[i]].set(xlim=(-0.05, 0.95), xticks=np.array(range(10)) / 10)
elif fig_i == 0:
axes[i][fig_row[i], fig_col[i]].set(xlim=(1, 17), xticks=np.array(range(2, 17)))
if i != 0:
axes[i][fig_row[i], fig_col[i]].set(ylim=(-0.05, 1.05), yticks=np.array(range(11)) / 10)
axes[i][fig_row[i], fig_col[i]].legend_.remove()
axes[i][fig_row[i], fig_col[i]].text(-0.2, 0.98, str(chr(ord('@')+fig_i+1)), fontfamily='sans-serif',
size=9, weight='bold', transform=axes[i][fig_row[i], fig_col[i]].transAxes)
# plots for sensitivity analyses on changing settings
for fig_i in range(n_fig, n_fig+3):
fig_row[0] = (fig_i + 1) // n_col[0]
fig_col[0] = (fig_i + 1) % n_col[0]
if fig_i == (n_fig + 1):
n_setting = 2
method = 2
name = 'oneforall'
xlabel = 'Homogeneity constraint for adopted strategies'
settings = ['Without', 'With']
elif fig_i == (n_fig + 2):
n_setting = 3
name = 'multi_methods'
xlabel = 'Surname inference approach'
settings = ['TMRCA-based', 'KNN', 'Linear regression']
elif fig_i == (n_fig + 0):
n_setting = 3
method = 2
name = 'multi_weight_distributions'
xlabel = 'Weight distribution of attributes'
settings = ['Entropy-based', 'Uniform', 'Highly biased']
folder_result_base = 'Results' + id_exp + '/Violin'
folder_result = folder_result_base + '_' + name + '/m' + str(method) + '/' # for alternative setting
folder_result_head = folder_result_base + '_over_confident/m' # for alternative inference methods
folder_result_head_input = folder_result_base + '_' + name + '/Alter_weight_' # for alternative weight distribution
folder_result_base += '/m' + str(method) + '/' # for default setting
col1 = []
rvs = []
for j in range(n_setting):
if fig_i == (n_fig + 2): # Surname inference approach
method = j + 2
folder = folder_result_head + str(method) + '/'
elif fig_i == (n_fig + 0) and j > 0: # Weight distribution of attributes
folder = folder_result_head_input + str(j) + '/m' + str(method) + '/'
elif j == 0:
folder = folder_result_base
else: # Homogeneity constraint for adopted strategies
folder = folder_result
if pruning == 1:
folder += 'pruning/'
if j == 1 and fig_i == (n_fig + 1): # Homogeneity constraint for adopted strategies
for i in range(3): # 3 game scenarios
scenario = i + 4
dataset = pd.read_pickle(folder + 'result_s' + str(scenario) + '.pickle')
col1.extend(dataset['defender_optimal'])
rvs.append(dataset['defender_optimal'])
continue # different data format
for i in range(3): # 3 game scenarios
scenario = i + 4
if scenario == 4 and fig_i == (n_fig + 0):
continue # no need to display the opt-in game
dataset = pd.read_pickle(folder + 'result_s' + str(scenario) + '.pickle')
data = np.array(dataset['defender_optimal'])
shaped_dataset = np.reshape(data, (n_iter, n_S))
av_dataset = np.mean(shaped_dataset, axis=1)
col1.extend(av_dataset)
rvs.append(av_dataset)
col2 = []
if fig_i != (n_fig + 0): # Not the setting on the weight distribution of attributes
for i in range(n_setting):
for j in range(3): # 3 game scenarios
col2.extend([scenario_name[j + 4] for k in range(n_iter)])
else:
for i in range(n_setting):
for j in range(2): # 2 game scenarios
col2.extend([scenario_name[j + 5] for k in range(n_iter)])
col3 = []
for i in range(n_setting):
if fig_i == (n_fig + 0):
n_game = 2
else:
n_game = 3
col3.extend([settings[i] for j in range(n_iter * n_game)])
dataset2 = pd.DataFrame({'Average payoff ($)': col1,
'Games': col2,
xlabel: col3})
if fig_i == (n_fig + 0):
colors1 = colors[5:7]
else:
colors1 = colors[4:7]
customPalette = sns.set_palette(sns.color_palette(colors1))
sns.violinplot(data=dataset2, x=xlabel, y='Average payoff ($)', hue='Games', scale='width', palette=customPalette,
ax=axes[0][fig_row[0], fig_col[0]])
axes[0][fig_row[0], fig_col[0]].legend_.remove()
axes[0][fig_row[0], fig_col[0]].text(- 0.2, 0.98, str(chr(ord('@') + fig_i + 1)), size=9, fontfamily='sans-serif',
transform=axes[0][fig_row[0], fig_col[0]].transAxes, weight='bold')
# Adjust, show and save each figure
for i in range(3):
fig[i].subplots_adjust(left=0, bottom=0, right=1, top=1, wspace=0, hspace=0)
ax_pos = axes[i][2, 2].get_position()
axes[i][2, 2].axis('off')
lines, labels = axes[i][0, 0].get_legend_handles_labels()
fig[i].legend(lines, labels, loc='upper left',
bbox_to_anchor=(ax_pos.x0 + 0.04, ax_pos.y1 - 0.02), borderaxespad=0.)
if i == 0:
lines, labels = axes[i][3, 2].get_legend_handles_labels()
fig[i].legend(lines, labels, loc='lower right', bbox_to_anchor=(ax_pos.x1 - 0.015, ax_pos.y0 + 0.01), borderaxespad=0.)
fig[i].show()
if i == 0:
pad_inch = 0.008
else:
pad_inch = 0.007
fig[i].savefig(folder_output + 'sensitivity_result_' + str(metric_name[i]) + '.png', bbox_inches='tight',
pad_inches=pad_inch, transparent=True, dpi=600)