質問編集履歴
1
時系列データに対するk近傍法のソースコードの追記
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -39,3 +39,183 @@
|
|
|
39
39
|
### 補足情報（FW/ツールのバージョンなど）
|
|
40
40
|
|
|
41
41
|
Pythonを使って実装しています。
|
|
42
|
+
|
|
43
|
+
|
|
44
|
+
|
|
45
|
+
```python
|
|
46
|
+
|
|
47
|
+
import pandas as pd
|
|
48
|
+
|
|
49
|
+
import numpy as np
|
|
50
|
+
|
|
51
|
+
import matplotlib.pyplot as plt
|
|
52
|
+
|
|
53
|
+
from sklearn.neighbors import NearestNeighbors
|
|
54
|
+
|
|
55
|
+
|
|
56
|
+
|
|
57
|
+
|
|
58
|
+
|
|
59
|
+
'''
|
|
60
|
+
|
|
61
|
+
dataをsize毎のスライス窓に分割
|
|
62
|
+
|
|
63
|
+
'''
|
|
64
|
+
|
|
65
|
+
|
|
66
|
+
|
|
67
|
+
def main():
|
|
68
|
+
|
|
69
|
+
df = pd.read_csv("20191121.csv")
|
|
70
|
+
|
|
71
|
+
|
|
72
|
+
|
|
73
|
+
# 余分なデータをDataFrameから削除
|
|
74
|
+
|
|
75
|
+
df = df.drop(['name', 'x_rad/s', 'y_rad/s', 'z_rad/s'], axis=1)
|
|
76
|
+
|
|
77
|
+
|
|
78
|
+
|
|
79
|
+
df = df.set_index('time')
|
|
80
|
+
|
|
81
|
+
|
|
82
|
+
|
|
83
|
+
#x, y, z軸加速度を視認化
|
|
84
|
+
|
|
85
|
+
df.plot().legend(loc='upper left')
|
|
86
|
+
|
|
87
|
+
|
|
88
|
+
|
|
89
|
+
# 前から2480件のx軸加速度を学習データ、その次の2479件をテストデータとする
|
|
90
|
+
|
|
91
|
+
# # df.iloc[2479] ---> 53845130
|
|
92
|
+
|
|
93
|
+
# df.iloc[2480] ---> 53845150
|
|
94
|
+
|
|
95
|
+
train_data = df.loc[:53845130, 'x_ags']
|
|
96
|
+
|
|
97
|
+
test_data = df.loc[53845150:, 'x_ags'].reset_index(drop=True)
|
|
98
|
+
|
|
99
|
+
|
|
100
|
+
|
|
101
|
+
# 窓幅
|
|
102
|
+
|
|
103
|
+
width = 30
|
|
104
|
+
|
|
105
|
+
|
|
106
|
+
|
|
107
|
+
# k近傍法のk
|
|
108
|
+
|
|
109
|
+
nk = 1
|
|
110
|
+
|
|
111
|
+
|
|
112
|
+
|
|
113
|
+
# 窓幅を使ってベクトルの集合を作成
|
|
114
|
+
|
|
115
|
+
train = embed(train_data, width)
|
|
116
|
+
|
|
117
|
+
test = embed(test_data, width)
|
|
118
|
+
|
|
119
|
+
|
|
120
|
+
|
|
121
|
+
# k近傍法でクラスタリング
|
|
122
|
+
|
|
123
|
+
neigh = NearestNeighbors(n_neighbors=nk)
|
|
124
|
+
|
|
125
|
+
neigh.fit(train)
|
|
126
|
+
|
|
127
|
+
|
|
128
|
+
|
|
129
|
+
# 距離を計算
|
|
130
|
+
|
|
131
|
+
d = neigh.kneighbors(test)[0]
|
|
132
|
+
|
|
133
|
+
|
|
134
|
+
|
|
135
|
+
# 距離の正規化
|
|
136
|
+
|
|
137
|
+
mx = np.max(d)
|
|
138
|
+
|
|
139
|
+
d = d / mx
|
|
140
|
+
|
|
141
|
+
|
|
142
|
+
|
|
143
|
+
# 訓練データ
|
|
144
|
+
|
|
145
|
+
plt.subplot(221)
|
|
146
|
+
|
|
147
|
+
plt.plot(train_data, label='Training')
|
|
148
|
+
|
|
149
|
+
plt.xlabel("Amplitude", fontsize=12)
|
|
150
|
+
|
|
151
|
+
plt.ylabel("Sample", fontsize=12)
|
|
152
|
+
|
|
153
|
+
plt.grid()
|
|
154
|
+
|
|
155
|
+
leg = plt.legend(loc=1, fontsize=15)
|
|
156
|
+
|
|
157
|
+
leg.get_frame().set_alpha(1)
|
|
158
|
+
|
|
159
|
+
|
|
160
|
+
|
|
161
|
+
# 異常度
|
|
162
|
+
|
|
163
|
+
plt.subplot(222)
|
|
164
|
+
|
|
165
|
+
plt.plot(d, label='d')
|
|
166
|
+
|
|
167
|
+
plt.xlabel("Amplitude", fontsize=12)
|
|
168
|
+
|
|
169
|
+
plt.ylabel("Sample", fontsize=12)
|
|
170
|
+
|
|
171
|
+
plt.grid()
|
|
172
|
+
|
|
173
|
+
leg = plt.legend(loc=1, fontsize=15)
|
|
174
|
+
|
|
175
|
+
leg.get_frame().set_alpha(1)
|
|
176
|
+
|
|
177
|
+
|
|
178
|
+
|
|
179
|
+
# 検証用データ
|
|
180
|
+
|
|
181
|
+
plt.subplot(223)
|
|
182
|
+
|
|
183
|
+
plt.plot(test_data, label='Test')
|
|
184
|
+
|
|
185
|
+
plt.xlabel("Amplitude", fontsize=12)
|
|
186
|
+
|
|
187
|
+
plt.ylabel("Sample", fontsize=12)
|
|
188
|
+
|
|
189
|
+
plt.grid()
|
|
190
|
+
|
|
191
|
+
leg = plt.legend(loc=1, fontsize=15)
|
|
192
|
+
|
|
193
|
+
leg.get_frame().set_alpha(1)
|
|
194
|
+
|
|
195
|
+
|
|
196
|
+
|
|
197
|
+
|
|
198
|
+
|
|
199
|
+
def embed(lst, dim):
|
|
200
|
+
|
|
201
|
+
emb = np.empty((0, dim), float)
|
|
202
|
+
|
|
203
|
+
for i in range(lst.size - dim + 1):
|
|
204
|
+
|
|
205
|
+
tmp = np.array(lst[i:i+dim])[::-1].reshape((1, -1))
|
|
206
|
+
|
|
207
|
+
emb = np.append(emb, tmp, axis=0)
|
|
208
|
+
|
|
209
|
+
return emb
|
|
210
|
+
|
|
211
|
+
|
|
212
|
+
|
|
213
|
+
if __name__ == '__main__':
|
|
214
|
+
|
|
215
|
+
main()
|
|
216
|
+
|
|
217
|
+
```
|
|
218
|
+
|
|
219
|
+
|
|
220
|
+
|
|
221
|
+
|