我正在尝试使用“ where”和“ count”功能的组合重新编码熊猫中的列值。期望的结果是在“ valence_median_split”列中从标记为“ Low_Valence”的行中选择200个随机行，并从标记为“ Low_Valence”的行中选择200个随机行。但是，这似乎不起作用。
这是df：
df.head()
Out[34]:
ID Category Num Vert_Horizon Description Fem_Valence_Mean \
0 Animals_001_h Animals 1 h Dead Stork 2.40
1 Animals_002_v Animals 2 v Lion 6.31
2 Animals_003_h Animals 3 h Snake 5.14
3 Animals_004_v Animals 4 v Wolf 4.55
4 Animals_005_h Animals 5 h Bat 5.29
Fem_Valence_SD Fem_Av_Ap_Mean Fem_Av/Ap_SD Arousal_Mean \
0 1.30 3.03 1.47 6.72
1 2.19 5.96 2.24 6.69
2 1.19 5.14 1.75 5.34
3 1.87 4.82 2.27 6.84
4 1.56 4.61 1.81 5.50
Luminance Contrast JPEG_size80 LABL LABA \
0 ... 126.05 68.45 263028 51.75 -0.39
1 ... 123.41 32.34 250208 52.39 10.63
2 ... 135.28 59.92 190887 55.45 0.25
3 ... 122.15 75.10 282350 49.84 3.82
4 ... 131.81 59.77 329325 54.26 -0.34
LABB Entropy Classification temp_selection valence_median_split
0 16.93 7.86 High Low_Valence
1 30.30 6.71 NaN High_Valence
2 4.41 7.83 High Low_Valence
3 1.36 7.69 High Low_Valence
4 -0.95 7.82 High Low_Valence
[5 rows x 35 columns]
这是我尝试过的：
df['temp_selection'] = ''
df['temp_selection'] = np.where(df['valence_median_split'] == 'Low_Valence', df['valence_median_split'].sample(n=200).reindex(df.index), 'Low')
df['temp_selection'] = np.where(df['valence_median_split'] == 'High_Valence', df['valence_median_split'].sample(n=200).reindex(df.index), 'High')
df.temp_selection.unique()
但是，结果表明这不起作用：
array(['High', nan, 'High_Valence'], dtype=object)
我想知道合并这些功能是否有错误。
以下是可重现的示例：
d = {'col1': [1, 2, 3, 4, 3, 3, 2, 2], 'col2': [1, 2, 3, 4, 3, 3, 2, 2]}
df = pd.DataFrame(data=d)
df['valence_median_split'] = ''
#Get median of valence
valence_median = df['col1'].median()
df['valence_median_split'] = np.where(df['col2'] < valence_median, 'Low_Valence', 'High_Valence')
df['temp_selection'] = ''
df['temp_selection'] = np.where(df['valence_median_split'] == 'Low_Valence', df['valence_median_split'].sample(n=2).reindex(df.index), 'Low')
df['temp_selection'] = np.where(df['valence_median_split'] == 'High_Valence', df['valence_median_split'].sample(n=2).reindex(df.index), 'High')
df
col1 col2 valence_median_split temp_selection
0 1 1 Low_Valence High
1 2 2 Low_Valence High
2 3 3 High_Valence High_Valence
3 4 4 High_Valence NaN
4 3 3 High_Valence NaN
5 3 3 High_Valence High_Valence
6 2 2 Low_Valence High
7 2 2 Low_Valence High
从上面的df中可以看出，在'temp_selection'中有一个'High_Valence'分类，不应存在，并且没有'Low'分类。
答案 0 :(得分：1)
想法是获取过滤后的数据ans的样本索引，而不是将numpy.select使用np.where的两倍：
low = df.loc[df['valence_median_split'] == 'Low_Valence',
'valence_median_split'].sample(n=2).index
high = df.loc[df['valence_median_split'] == 'High_Valence',
'valence_median_split'].sample(n=2).index
df['temp_selection'] = np.select([df.index.isin(low), df.index.isin(high)],
['Low', 'High'], default=np.nan)
或者：
df['temp_selection'] = np.where(df.index.isin(low), 'Low',
np.where(df.index.isin(high), 'High', np.nan))
print (df)
col1 col2 valence_median_split temp_selection
0 1 1 Low_Valence nan
1 2 2 Low_Valence Low
2 3 3 High_Valence nan
3 4 4 High_Valence nan
4 3 3 High_Valence High
5 3 3 High_Valence High
6 2 2 Low_Valence nan
7 2 2 Low_Valence Low
或者：
df.loc[low, 'temp_selection'] = 'Low'
df.loc[high, 'temp_selection'] = 'High'
print (df)
col1 col2 valence_median_split temp_selection
0 1 1 Low_Valence NaN
1 2 2 Low_Valence Low
2 3 3 High_Valence NaN
3 4 4 High_Valence NaN
4 3 3 High_Valence High
5 3 3 High_Valence High
6 2 2 Low_Valence NaN
7 2 2 Low_Valence Low
另一个想法是使用numpy.random.choice：
low = np.random.choice(df.index[df['valence_median_split'] == 'Low_Valence'], size=2)
high = np.random.choice(df.index[df['valence_median_split']== 'High_Valence'], size=2)
df.loc[low, 'temp_selection'] = 'Low'
df.loc[high, 'temp_selection'] = 'High'