質問編集履歴
1
文章の修正
test
CHANGED
@@ -1 +1 @@
|
|
1
|
-
|
1
|
+
エラーメッセージの意味を知りたい
|
test
CHANGED
@@ -1,8 +1,8 @@
|
|
1
1
|
### 前提・実現したいこと
|
2
2
|
|
3
|
+
エラーメッセージの意味を知りたい
|
3
4
|
|
4
5
|
|
5
|
-
SVMの多クラス分類における分離超平面を描きたい
|
6
6
|
|
7
7
|
### 発生している問題・エラーメッセージ
|
8
8
|
|
@@ -14,520 +14,4 @@
|
|
14
14
|
|
15
15
|
```
|
16
16
|
|
17
|
-
|
18
|
-
|
19
|
-
### 該当のソースコード
|
20
|
-
|
21
|
-
|
22
|
-
|
23
|
-
```Python
|
24
|
-
|
25
|
-
import pandas as pd
|
26
|
-
|
27
|
-
import numpy as np
|
28
|
-
|
29
|
-
#import seaborn as sns
|
30
|
-
|
31
|
-
#sns.set_style("darkgrid")
|
32
|
-
|
33
|
-
#3次元プロットするためのモジュール
|
34
|
-
|
35
|
-
from mpl_toolkits.mplot3d import Axes3D
|
36
|
-
|
37
|
-
import matplotlib.pyplot as plt
|
38
|
-
|
39
|
-
from matplotlib.colors import LinearSegmentedColormap
|
40
|
-
|
41
|
-
|
42
|
-
|
43
|
-
|
44
|
-
|
45
|
-
def generate_cmap(colors, num):
|
46
|
-
|
47
|
-
|
48
|
-
|
49
|
-
"""自分で定義したカラーマップを返す"""
|
50
|
-
|
51
|
-
|
52
|
-
|
53
|
-
values = range(num)
|
54
|
-
|
55
|
-
|
56
|
-
|
57
|
-
vmax = np.ceil(np.max(values))
|
58
|
-
|
59
|
-
|
60
|
-
|
61
|
-
color_list = []
|
62
|
-
|
63
|
-
|
64
|
-
|
65
|
-
for v, c in zip(values, colors):
|
66
|
-
|
67
|
-
|
68
|
-
|
69
|
-
color_list.append( ( v/ vmax, c) )
|
70
|
-
|
71
|
-
|
72
|
-
|
73
|
-
return LinearSegmentedColormap.from_list('custom_cmap', color_list)
|
74
|
-
|
75
|
-
|
76
|
-
|
77
|
-
def plot_classifier2(cls, sc, X_train, y_train, X_test, y_test, y_pred, clis_f, CL_lis, MK_lis, St, num):
|
78
|
-
|
79
|
-
|
80
|
-
|
81
|
-
|
82
|
-
|
83
|
-
num1 = len(y_test)
|
84
|
-
|
85
|
-
|
86
|
-
|
87
|
-
dum1 = 0
|
88
|
-
|
89
|
-
|
90
|
-
|
91
|
-
for i in range(num1):
|
92
|
-
|
93
|
-
|
94
|
-
|
95
|
-
if(y_test[i] == y_pred[i]):
|
96
|
-
|
97
|
-
|
98
|
-
|
99
|
-
dum1 +=1
|
100
|
-
|
101
|
-
print("accuracy rate (percent):")
|
102
|
-
|
103
|
-
|
104
|
-
|
105
|
-
accu = (1.0*dum1)/(1.0*num1)*100.0
|
106
|
-
|
107
|
-
|
108
|
-
|
109
|
-
print(accu)
|
110
|
-
|
111
|
-
|
112
|
-
|
113
|
-
fig = plt.figure()
|
114
|
-
|
115
|
-
ax = Axes3D(fig)
|
116
|
-
|
117
|
-
|
118
|
-
|
119
|
-
x1_min, x1_max = -2.5, 2.5
|
120
|
-
|
121
|
-
|
122
|
-
|
123
|
-
x2_min, x2_max = -2.5, 2.5
|
124
|
-
|
125
|
-
"""resolutionは境界線の滑らかさ"""
|
126
|
-
|
127
|
-
resolution = 0.02
|
128
|
-
|
129
|
-
|
130
|
-
|
131
|
-
xx1, xx2= np.meshgrid(np.arange(x1_min, x1_max, resolution),
|
132
|
-
|
133
|
-
|
134
|
-
|
135
|
-
np.arange(x2_min, x2_max, resolution))
|
136
|
-
|
137
|
-
#print(xx1)
|
138
|
-
|
139
|
-
Xp = np.array([xx1.ravel(), xx2.ravel()]).T
|
140
|
-
|
141
|
-
#print(Xp)
|
142
|
-
|
143
|
-
|
144
|
-
|
145
|
-
"""scは標準化(後述で記載)"""
|
146
|
-
|
147
|
-
Xp1 = sc.transform(Xp)
|
148
|
-
|
149
|
-
|
150
|
-
|
151
|
-
"""clsの意味は?"""
|
152
|
-
|
153
|
-
Z = cls.predict(Xp1)
|
154
|
-
|
155
|
-
|
156
|
-
|
157
|
-
clis = []
|
158
|
-
|
159
|
-
"""uniqueは重複要素の削除"""
|
160
|
-
|
161
|
-
yc = np.unique(Z)
|
162
|
-
|
163
|
-
|
164
|
-
|
165
|
-
yt = np.unique(y_train)
|
166
|
-
|
167
|
-
|
168
|
-
|
169
|
-
"""appendは要素の追加"""
|
170
|
-
|
171
|
-
for i in range(len(np.unique(yc))):
|
172
|
-
|
173
|
-
|
174
|
-
|
175
|
-
for j in range(len(np.unique(yt))):
|
176
|
-
|
177
|
-
|
178
|
-
|
179
|
-
if yc[i] == yt[j]:
|
180
|
-
|
181
|
-
|
182
|
-
|
183
|
-
clis.append(clis_f[j])
|
184
|
-
|
185
|
-
|
186
|
-
|
187
|
-
|
188
|
-
|
189
|
-
cm = generate_cmap(clis, len(np.unique(Z)))
|
190
|
-
|
191
|
-
#reshapeh列数自動で形状の変更
|
192
|
-
|
193
|
-
Z1 = Z.reshape(xx1.shape)
|
194
|
-
|
195
|
-
ax.contourf(xx1, xx2, Z1, alpha=0.3, cmap=cm)
|
196
|
-
|
197
|
-
#reshapeh列数自動で形状の変
|
198
|
-
|
199
|
-
|
200
|
-
|
201
|
-
ax.set_xlabel("V1-V2 pair density (normalized)")
|
202
|
-
|
203
|
-
ax.set_ylabel("V1-V6 pair density (normalized)")
|
204
|
-
|
205
|
-
ax.set_zlabel("V1-V4 pair density (normalized)")
|
206
|
-
|
207
|
-
|
208
|
-
|
209
|
-
for idx in range(len(np.unique(y0))):
|
210
|
-
|
211
|
-
ax.plot(X_train[y_train==idx+1, 0], X_train[y_train==idx+1, 1], X_train[y_train==idx+1, 2],
|
212
|
-
|
213
|
-
color=CL_lis[idx], marker=MK_lis[idx], label=LB_lis[idx], linestyle='None')
|
214
|
-
|
215
|
-
#ZZ = -(w[idx:, 0] * xx1 + w[idx, 1] * xx2 + b)/ w[idx, 2]
|
216
|
-
|
217
|
-
|
218
|
-
|
219
|
-
#ax.plot_surface(xx1, xx2, ZZ, alpha = 0.3, cmap = cm)
|
220
|
-
|
221
|
-
|
222
|
-
|
223
|
-
ax.plot(X_test[y_test==y_pred,0], X_test[y_test==y_pred,1], X_test[y_test==y_pred,2],
|
224
|
-
|
225
|
-
|
226
|
-
|
227
|
-
color='black', marker='o', label="test(true)", linestyle='None')
|
228
|
-
|
229
|
-
|
230
|
-
|
231
|
-
ax.plot(X_test[y_test!=y_pred,0], X_test[y_test!=y_pred,1], X_test[y_test!=y_pred,2],
|
232
|
-
|
233
|
-
|
234
|
-
|
235
|
-
color='black', marker='x', label="test(false)", linestyle='None')
|
236
|
-
|
237
|
-
|
238
|
-
|
239
|
-
|
240
|
-
|
241
|
-
plt.legend(loc='upper left')
|
242
|
-
|
243
|
-
|
244
|
-
|
245
|
-
plt.show()
|
246
|
-
|
247
|
-
|
248
|
-
|
249
|
-
return accu
|
250
|
-
|
251
|
-
|
252
|
-
|
253
|
-
|
254
|
-
|
255
|
-
df0 = pd.read_csv('C:\sample\xxxxx.csv', header = None)
|
256
|
-
|
257
|
-
|
258
|
-
|
259
|
-
df0.columns = ['V1V2', 'V1V35', 'V1V6', 'V1V4', 'KAM','class label']
|
260
|
-
|
261
|
-
df0.head()
|
262
|
-
|
263
|
-
|
264
|
-
|
265
|
-
y0 = df0.iloc[0:, 5].values
|
266
|
-
|
267
|
-
#1列目
|
268
|
-
|
269
|
-
X0_V1V2 = np.copy(df0.iloc[0:, 0].values)
|
270
|
-
|
271
|
-
#2列目
|
272
|
-
|
273
|
-
X0_V1V35 = np.copy(df0.iloc[0:, 1].values)
|
274
|
-
|
275
|
-
#3列目
|
276
|
-
|
277
|
-
X0_V1V6 = np.copy(df0.iloc[0:, 2].values)
|
278
|
-
|
279
|
-
#4列目
|
280
|
-
|
281
|
-
X0_V1V4 = np.copy(df0.iloc[0:, 3].values)
|
282
|
-
|
283
|
-
#5列目
|
284
|
-
|
285
|
-
X0_KAM = np.copy(df0.iloc[0:, 4].values)
|
286
|
-
|
287
|
-
|
288
|
-
|
289
|
-
X0 = df0.iloc[0:, [0,2,3]].values
|
290
|
-
|
291
|
-
|
292
|
-
|
293
|
-
from sklearn.model_selection import train_test_split
|
294
|
-
|
295
|
-
|
296
|
-
|
297
|
-
X_train, X_test, y_train, y_test = train_test_split(X0, y0, test_size=0.2, random_state=0)
|
298
|
-
|
299
|
-
|
300
|
-
|
301
|
-
from sklearn.preprocessing import StandardScaler
|
302
|
-
|
303
|
-
#標準化
|
304
|
-
|
305
|
-
sc = StandardScaler()
|
306
|
-
|
307
|
-
|
308
|
-
|
309
|
-
X_train_std = sc.fit_transform(X_train)
|
310
|
-
|
311
|
-
|
312
|
-
|
313
|
-
X_test_std = sc.transform(X_test)
|
314
|
-
|
315
|
-
|
316
|
-
|
317
|
-
from sklearn.svm import SVC
|
318
|
-
|
319
|
-
|
320
|
-
|
321
|
-
svm = SVC(kernel='linear', C=1, random_state=0)
|
322
|
-
|
323
|
-
|
324
|
-
|
325
|
-
svm.fit(X_train_std,y_train)
|
326
|
-
|
327
|
-
|
328
|
-
|
329
|
-
y_pred = svm.predict(X_test_std)
|
330
|
-
|
331
|
-
#plot_classifier(svm, sc, X0, y0, y_train, clis_f, CL_lis, MK_lis, St)
|
332
|
-
|
333
|
-
|
334
|
-
|
335
|
-
accu_t = 0.0
|
336
|
-
|
337
|
-
|
338
|
-
|
339
|
-
kk = 5
|
340
|
-
|
341
|
-
|
342
|
-
|
343
|
-
z = 0.0
|
344
|
-
|
345
|
-
|
346
|
-
|
347
|
-
CL_lis = ['red','blue','green','grey','cyan','pink','salmon']
|
348
|
-
|
349
|
-
|
350
|
-
|
351
|
-
MK_lis = ['o','v','^','+','D','o', 'D']
|
352
|
-
|
353
|
-
|
354
|
-
|
355
|
-
LB_lis = ['T1','T2','T3/T4/T5','T6','T7','T8', 'H']
|
356
|
-
|
357
|
-
|
358
|
-
|
359
|
-
St = ""
|
360
|
-
|
361
|
-
|
362
|
-
|
363
|
-
clis_f = ['red','blue','green','grey','cyan','yellow']
|
364
|
-
|
365
|
-
|
366
|
-
|
367
|
-
for i in range(kk):
|
368
|
-
|
369
|
-
|
370
|
-
|
371
|
-
X_train, X_test, y_train, y_test = train_test_split(X0, y0, test_size=0.2, random_state=i)
|
372
|
-
|
373
|
-
|
374
|
-
|
375
|
-
sc = StandardScaler()
|
376
|
-
|
377
|
-
|
378
|
-
|
379
|
-
X_train_std = sc.fit_transform(X_train)
|
380
|
-
|
381
|
-
|
382
|
-
|
383
|
-
X_test_std = sc.transform(X_test)
|
384
|
-
|
385
|
-
|
386
|
-
|
387
|
-
svm = SVC(kernel='linear', C=10.0, random_state=i)
|
388
|
-
|
389
|
-
|
390
|
-
|
391
|
-
svm.fit(X_train_std,y_train)
|
392
|
-
|
393
|
-
|
394
|
-
|
395
|
-
y_pred = svm.predict(X_test_std)
|
396
|
-
|
397
|
-
|
398
|
-
|
399
|
-
plot_classifier2(svm, sc, X_train, y_train, X_test, y_test, y_pred, clis_f, CL_lis, MK_lis, St, num=i)
|
400
|
-
|
401
|
-
|
402
|
-
|
403
|
-
num1 = len(y_test)
|
404
|
-
|
405
|
-
|
406
|
-
|
407
|
-
dum1 = 0
|
408
|
-
|
409
|
-
|
410
|
-
|
411
|
-
for i in range(num1):
|
412
|
-
|
413
|
-
|
414
|
-
|
415
|
-
if(y_test[i] == y_pred[i]):
|
416
|
-
|
417
|
-
|
418
|
-
|
419
|
-
dum1 +=1
|
420
|
-
|
421
|
-
|
422
|
-
|
423
|
-
accu_t = (1.0*dum1)/(1.0*num1)*100.0
|
424
|
-
|
425
|
-
|
426
|
-
|
427
|
-
#print("accu. = ",accu_t)
|
428
|
-
|
429
|
-
|
430
|
-
|
431
|
-
z += accu_t
|
432
|
-
|
433
|
-
|
434
|
-
|
435
|
-
print("total accu. = ",z / kk)
|
436
|
-
|
437
|
-
|
438
|
-
|
439
|
-
|
440
|
-
|
441
|
-
|
442
|
-
|
443
|
-
import numpy as np
|
444
|
-
|
445
|
-
X = X0_V1V2
|
446
|
-
|
447
|
-
Y = X0_V1V6
|
448
|
-
|
449
|
-
Z = X0_V1V4
|
450
|
-
|
451
|
-
|
452
|
-
|
453
|
-
|
454
|
-
|
455
|
-
#seabornでグラフをきれいにしたいだけのコード
|
456
|
-
|
457
|
-
|
458
|
-
|
459
|
-
#sns.set_style("darkgrid")
|
460
|
-
|
461
|
-
|
462
|
-
|
463
|
-
|
464
|
-
|
465
|
-
|
466
|
-
|
467
|
-
|
468
|
-
|
469
|
-
#グラフの枠を作っていく
|
470
|
-
|
471
|
-
fig = plt.figure()
|
472
|
-
|
473
|
-
ax = Axes3D(fig)
|
474
|
-
|
475
|
-
|
476
|
-
|
477
|
-
#軸にラベルを付けたいときは書く
|
478
|
-
|
479
|
-
ax.set_xlabel("V1-V2 pair density (normalized)")
|
480
|
-
|
481
|
-
ax.set_ylabel("V1-V6 pair density (normalized)")
|
482
|
-
|
483
|
-
ax.set_zlabel("V1-V4 pair density (normalized)")
|
484
|
-
|
485
|
-
|
486
|
-
|
487
|
-
#.plotで描画
|
488
|
-
|
489
|
-
#linestyle='None'にしないと初期値では線が引かれるが、3次元の散布図だと大抵ジャマになる
|
490
|
-
|
491
|
-
#markerは無難に丸
|
492
|
-
|
493
|
-
for idx in range(len(np.unique(y0))):
|
494
|
-
|
495
|
-
ax.plot(X[y0==idx+1], Y[y0==idx+1], Z[y0==idx+1],
|
496
|
-
|
497
|
-
color=CL_lis[idx], marker=MK_lis[idx], label=LB_lis[idx], linestyle='None')
|
498
|
-
|
499
|
-
|
500
|
-
|
501
|
-
'''ax.plot(X_test[y_test==y_pred,0], X_test[y_test==y_pred,1],
|
502
|
-
|
503
|
-
|
504
|
-
|
505
|
-
color='black', marker='o', label="test(true)", linestyle='None')
|
506
|
-
|
507
|
-
|
508
|
-
|
509
|
-
ax.plot(X_test[y_test!=y_pred,0], X_test[y_test!=y_pred,1],
|
510
|
-
|
511
|
-
|
512
|
-
|
513
|
-
color='black', marker='x', label="test(false)", linestyle='None')'''
|
514
|
-
|
515
|
-
plt.legend(loc='upper left')
|
516
|
-
|
517
|
-
#最後に.show()を書いてグラフ表示
|
518
|
-
|
519
|
-
plt.show()
|
520
|
-
|
521
|
-
```
|
522
|
-
|
523
|
-
|
524
|
-
|
525
|
-
### 試したこと
|
526
|
-
|
527
|
-
|
528
|
-
|
529
|
-
おそらくax.plot_surfaceを使うと思うのですが, うまくいかなさそうなので, https://matplotlib.org/mpl_toolkits/mplot3d/tutorial.htmlのAxes3D.contourfを使ってそれぞれの平面での境界線を描こうと思い実行したところ上記のエラーが発生しました. エラーメッセージの意味だけでも教えていただけると幸いです。
|
530
|
-
|
531
|
-
|
532
|
-
|
533
17
|
### 補足情報(ツールのバージョンなど)
|