回答編集履歴
1
d
test
CHANGED
|
@@ -93,3 +93,91 @@
|
|
|
93
93
|
|
|
94
94
|
|
|
95
95
|
|
|
96
|
+
|
|
97
|
+
|
|
98
|
+
|
|
99
|
+
## 数でなく、割合で混合行列を表示したい場合
|
|
100
|
+
|
|
101
|
+
|
|
102
|
+
|
|
103
|
+
```python
|
|
104
|
+
|
|
105
|
+
import matplotlib.pyplot as plt
|
|
106
|
+
|
|
107
|
+
import numpy as np
|
|
108
|
+
|
|
109
|
+
import pandas as pd
|
|
110
|
+
|
|
111
|
+
import seaborn as sns
|
|
112
|
+
|
|
113
|
+
from sklearn.metrics import confusion_matrix
|
|
114
|
+
|
|
115
|
+
from sklearn.neighbors import KNeighborsClassifier
|
|
116
|
+
|
|
117
|
+
|
|
118
|
+
|
|
119
|
+
sns.set()
|
|
120
|
+
|
|
121
|
+
|
|
122
|
+
|
|
123
|
+
df = pd.read_csv('data.csv')
|
|
124
|
+
|
|
125
|
+
|
|
126
|
+
|
|
127
|
+
# 学習データ、テストデータに分割する。
|
|
128
|
+
|
|
129
|
+
train_x, test_x, train_y, test_y = MS.train_test_split(
|
|
130
|
+
|
|
131
|
+
df.drop('result', axis=1), df['result'], test_size=0.2)
|
|
132
|
+
|
|
133
|
+
|
|
134
|
+
|
|
135
|
+
clf = KNeighborsClassifier(n_neighbors=3)
|
|
136
|
+
|
|
137
|
+
clf.fit(train_x, train_y)
|
|
138
|
+
|
|
139
|
+
|
|
140
|
+
|
|
141
|
+
# テストデータを推論する。
|
|
142
|
+
|
|
143
|
+
pred_y = clf.predict(test_x)
|
|
144
|
+
|
|
145
|
+
|
|
146
|
+
|
|
147
|
+
# 混合行列を作成する。
|
|
148
|
+
|
|
149
|
+
cm = confusion_matrix(test_y, pred_y)
|
|
150
|
+
|
|
151
|
+
cm = cm / cm.sum()
|
|
152
|
+
|
|
153
|
+
|
|
154
|
+
|
|
155
|
+
# 混合行列を描画する。
|
|
156
|
+
|
|
157
|
+
def print_confusion_matrix(confusion_matrix, class_names):
|
|
158
|
+
|
|
159
|
+
heatmap = sns.heatmap(
|
|
160
|
+
|
|
161
|
+
confusion_matrix, xticklabels=class_names, yticklabels=class_names,
|
|
162
|
+
|
|
163
|
+
annot=True, fmt='.2%', cbar=False, square=True, cmap='YlGnBu')
|
|
164
|
+
|
|
165
|
+
plt.ylabel('True label')
|
|
166
|
+
|
|
167
|
+
plt.xlabel('Predicted label')
|
|
168
|
+
|
|
169
|
+
plt.show()
|
|
170
|
+
|
|
171
|
+
|
|
172
|
+
|
|
173
|
+
|
|
174
|
+
|
|
175
|
+
labels = ['class {}'.format(i) for i in df['result'].unique()]
|
|
176
|
+
|
|
177
|
+
print_confusion_matrix(cm, labels)
|
|
178
|
+
|
|
179
|
+
```
|
|
180
|
+
|
|
181
|
+
|
|
182
|
+
|
|
183
|
+
|