回答編集履歴
1
いろんなコードを試行してみました
test
CHANGED
|
@@ -1,6 +1,10 @@
|
|
|
1
|
+
その1：質問のコードを直したもの
|
|
2
|
+
|
|
3
|
+
|
|
4
|
+
|
|
1
|
-
```python
|
|
5
|
+
```python
|
|
2
|
-
|
|
6
|
+
|
|
3
|
-
df['Season'].astype(str).str.cat([
|
|
7
|
+
df["ID"] = df['Season'].astype(str).str.cat([
|
|
4
8
|
|
|
5
9
|
df[['WTeamID', 'LTeamID']].min(1).astype(str),
|
|
6
10
|
|
|
@@ -10,7 +14,39 @@
|
|
|
10
14
|
|
|
11
15
|
|
|
12
16
|
|
|
17
|
+
その2：より簡単なコード
|
|
18
|
+
|
|
19
|
+
|
|
20
|
+
|
|
21
|
+
```python
|
|
22
|
+
|
|
23
|
+
df["ID"] = df['Season'].astype(str).str.cat(
|
|
24
|
+
|
|
25
|
+
np.sort(df[['LTeamID', 'WTeamID']].to_numpy()).astype(str), '_')
|
|
26
|
+
|
|
27
|
+
```
|
|
28
|
+
|
|
29
|
+
|
|
30
|
+
|
|
31
|
+
その3：リストを使う方法
|
|
32
|
+
|
|
33
|
+
|
|
34
|
+
|
|
35
|
+
```python
|
|
36
|
+
|
|
37
|
+
arr_s = df['Season'].tolist()
|
|
38
|
+
|
|
39
|
+
arr_lw = np.sort(df[['LTeamID', 'WTeamID']].to_numpy()).tolist()
|
|
40
|
+
|
|
41
|
+
df["ID"] = pd.Series([f'{s}_{lw[0]}_{lw[1]}' for s, lw in zip(arr_s, arr_lw)],
|
|
42
|
+
|
|
43
|
+
index=df.index)
|
|
44
|
+
|
|
45
|
+
```
|
|
46
|
+
|
|
47
|
+
|
|
48
|
+
|
|
13
|
-
## 解説
|
|
49
|
+
## 解説（その1）
|
|
14
50
|
|
|
15
51
|
|
|
16
52
|
|
|
@@ -176,13 +212,13 @@
|
|
|
176
212
|
|
|
177
213
|
```python
|
|
178
214
|
|
|
179
|
-
In [2
|
|
215
|
+
In [26]: df['Season'].astype(str).str.cat([
|
|
180
216
|
|
|
181
217
|
2 df[['WTeamID', 'LTeamID']].min(1).astype(str),
|
|
182
218
|
|
|
183
219
|
3 df[['WTeamID', 'LTeamID']].max(1).astype(str)], '_')
|
|
184
220
|
|
|
185
|
-
Out[2
|
|
221
|
+
Out[26]:
|
|
186
222
|
|
|
187
223
|
1916 2015_1214_1264
|
|
188
224
|
|
|
@@ -209,3 +245,161 @@
|
|
|
209
245
|
Name: Season, dtype: object
|
|
210
246
|
|
|
211
247
|
```
|
|
248
|
+
|
|
249
|
+
|
|
250
|
+
|
|
251
|
+
## 解説（その2）
|
|
252
|
+
|
|
253
|
+
|
|
254
|
+
|
|
255
|
+
今回の場合、数値を比較して「小さい方_大きい方」という部分がありますが、この部分は`np.sort()`を使うって行ごとに「小さい→大きい」順に並べ替えした配列を作るのが簡便です。
|
|
256
|
+
|
|
257
|
+
つまり、
|
|
258
|
+
|
|
259
|
+
|
|
260
|
+
|
|
261
|
+
```python
|
|
262
|
+
|
|
263
|
+
In [5]: arr = np.sort(df[['LTeamID', 'WTeamID']].to_numpy())
|
|
264
|
+
|
|
265
|
+
2 arr
|
|
266
|
+
|
|
267
|
+
Out[5]:
|
|
268
|
+
|
|
269
|
+
array([[1214, 1264],
|
|
270
|
+
|
|
271
|
+
[1140, 1279],
|
|
272
|
+
|
|
273
|
+
[1129, 1173],
|
|
274
|
+
|
|
275
|
+
[1316, 1352],
|
|
276
|
+
|
|
277
|
+
[1112, 1411],
|
|
278
|
+
|
|
279
|
+
... ...
|
|
280
|
+
|
|
281
|
+
[1181, 1211],
|
|
282
|
+
|
|
283
|
+
[1257, 1277],
|
|
284
|
+
|
|
285
|
+
[1181, 1277],
|
|
286
|
+
|
|
287
|
+
[1246, 1458],
|
|
288
|
+
|
|
289
|
+
[1181, 1458]], dtype=int64)
|
|
290
|
+
|
|
291
|
+
|
|
292
|
+
|
|
293
|
+
In [6]: df['Season'].astype(str).str.cat(arr.astype(str), '_')
|
|
294
|
+
|
|
295
|
+
Out[6]:
|
|
296
|
+
|
|
297
|
+
1916 2015_1214_1264
|
|
298
|
+
|
|
299
|
+
1917 2015_1140_1279
|
|
300
|
+
|
|
301
|
+
1918 2015_1129_1173
|
|
302
|
+
|
|
303
|
+
1919 2015_1316_1352
|
|
304
|
+
|
|
305
|
+
1920 2015_1112_1411
|
|
306
|
+
|
|
307
|
+
... ...
|
|
308
|
+
|
|
309
|
+
1978 2015_1181_1211
|
|
310
|
+
|
|
311
|
+
1979 2015_1257_1277
|
|
312
|
+
|
|
313
|
+
1980 2015_1181_1277
|
|
314
|
+
|
|
315
|
+
1981 2015_1246_1458
|
|
316
|
+
|
|
317
|
+
1982 2015_1181_1458
|
|
318
|
+
|
|
319
|
+
Name: Season, dtype: object
|
|
320
|
+
|
|
321
|
+
```
|
|
322
|
+
|
|
323
|
+
|
|
324
|
+
|
|
325
|
+
## 解説（その3）
|
|
326
|
+
|
|
327
|
+
|
|
328
|
+
|
|
329
|
+
pandasの`str`関連メソッドは実際にはとても動作が遅いです。標準のリストになおしてforループした方が速い場合があります。
|
|
330
|
+
|
|
331
|
+
|
|
332
|
+
|
|
333
|
+
```python
|
|
334
|
+
|
|
335
|
+
In [11]: arr_s = df['Season'].tolist()
|
|
336
|
+
|
|
337
|
+
2 arr_lw = np.sort(df[['LTeamID', 'WTeamID']].to_numpy()).tolist()
|
|
338
|
+
|
|
339
|
+
3 id_list = [f'{s}_{lw[0]}_{lw[1]}' for s, lw in zip(arr_s, arr_lw)]
|
|
340
|
+
|
|
341
|
+
4 id_list
|
|
342
|
+
|
|
343
|
+
Out[11]:
|
|
344
|
+
|
|
345
|
+
['2015_1214_1264',
|
|
346
|
+
|
|
347
|
+
'2015_1140_1279',
|
|
348
|
+
|
|
349
|
+
'2015_1129_1173',
|
|
350
|
+
|
|
351
|
+
'2015_1316_1352',
|
|
352
|
+
|
|
353
|
+
'2015_1112_1411',
|
|
354
|
+
|
|
355
|
+
'2015_1181_1211',
|
|
356
|
+
|
|
357
|
+
'2015_1257_1277',
|
|
358
|
+
|
|
359
|
+
'2015_1181_1277',
|
|
360
|
+
|
|
361
|
+
'2015_1246_1458',
|
|
362
|
+
|
|
363
|
+
'2015_1181_1458']
|
|
364
|
+
|
|
365
|
+
|
|
366
|
+
|
|
367
|
+
In [12]: pd.Series(id_list, index=df.index)
|
|
368
|
+
|
|
369
|
+
Out[12]:
|
|
370
|
+
|
|
371
|
+
1916 2015_1214_1264
|
|
372
|
+
|
|
373
|
+
1917 2015_1140_1279
|
|
374
|
+
|
|
375
|
+
1918 2015_1129_1173
|
|
376
|
+
|
|
377
|
+
1919 2015_1316_1352
|
|
378
|
+
|
|
379
|
+
1920 2015_1112_1411
|
|
380
|
+
|
|
381
|
+
1978 2015_1181_1211
|
|
382
|
+
|
|
383
|
+
1979 2015_1257_1277
|
|
384
|
+
|
|
385
|
+
1980 2015_1181_1277
|
|
386
|
+
|
|
387
|
+
1981 2015_1246_1458
|
|
388
|
+
|
|
389
|
+
1982 2015_1181_1458
|
|
390
|
+
|
|
391
|
+
dtype: object
|
|
392
|
+
|
|
393
|
+
```
|
|
394
|
+
|
|
395
|
+
|
|
396
|
+
|
|
397
|
+
## 処理速度の比較
|
|
398
|
+
|
|
399
|
+
|
|
400
|
+
|
|
401
|
+
|
|
402
|
+
|
|
403
|
+
|
|
404
|
+
|
|
405
|
+
その３（リストを使う方法）が最も速いようです。@can110氏の方法（`df.apply`を使った方法）はデータ数が少ないときは次いで速いですが、行数が増えるとかなり遅くなります。
|