熊猫，根据行中的值创建列

Posted 2023-03-17

【中文标题】熊猫，根据行中的值创建列

【英文标题】：Pandas, create columns based on value in row

【发布时间】：2022-01-01 01:29:24

【问题描述】：

我有以下数据框（这只是一行）：

X0  X1  X2  X3  X4  X5  X6  X8
32  23  17  0   3   24  9   14

我想将 X8 列中的值 14 映射到以下列表mylist：

[ 0.06405287, -0.1176078 , -0.06206927,  0.08389127, -0.18036067,
        0.35158703, -0.0928449 , -0.0974429 , -0.06705306, -0.17196381,
       -0.03776502,  0.09204011,  0.47813812,  0.16258538,  0.2699648 ,
        0.07496626, -0.09791522, -0.31499937, -0.24898018,  0.06126055,
        0.13187763,  0.21042736, -0.1585868 ,  0.08355565, -0.13935572,
        0.12408883,  0.2043313 , -0.12544186, -0.09223691,  0.00720569]

为了这个问题，这里有一个创建一个数据帧的 for 循环。

col = 
col['X8'] =  #
col['X8']['14']=  # just creating the double nest for the sake of the example, this would of course be 3 for loops because I need to match each unique value in each column to its corresponding list of 30 values
for i, x in mylist:
  col['X8']['14'][i] = x 

colx814 = pd.DataFrame.from_dict(col['X8']['14'],orient='index').T 

现在我有一个如下所示的数据框：

0   1   2   3   4   5   6   7   8   9   10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29
0.064053    -0.117608   -0.062069   0.083891    -0.180361   0.351587    -0.092845   -0.097443   -0.067053   -0.171964   -0.037765   0.09204 0.478138    0.162585    0.269965    0.074966    -0.097915   -0.314999   -0.24898    0.061261    0.131878    0.210427    -0.158587   0.083556    -0.139356   0.124089    0.204331    -0.125442   -0.092237   0.007206

如何将其加入我的原始数据框？我的猜测是map+apply+concat，但我不确定如何有效地组合它们。另一种解决方案是将每一列复制 30 次，然后对每一列进行迭代+替换？我搜索并找不到这个问题的答案，所以我想我会发布它以防有人有答案，如果没有自己回答。这就是我要找的：

X0  X1  X2  X3  X4  X5  X6  X8 X8_0 X8_1 X8_2 ...etc
32  23  17  0   3   24  9   14 mylist[0] mylist[1] mylist[2] ...etc

当然，每一列都有自己的 30 个 XN_n 列。

【参考方案1】：

weights = 
for index, x in enumerate(encoded.columns):
  weights[x] = 
  for id, val in enumerate(encoded[x].unique()):
    weights[x][val] = 
    for weightid, weightval in enumerate(model_full.get_layer(embeddings[index]).get_weights()[0][id]):
      weights[x][val][weightid] = weightval

mappedembeddings = pd.DataFrame()
encodedindex = []
for row in encoded.iterrows():
  encodedindex.append(row[0])
  df0 = pd.DataFrame()
  for k, v in row[1].to_dict().items():
    names = []
    for z in weights[k][v].keys():
      names.append(str(k)+'_'+str(z))

    tempdf = pd.DataFrame([weights[k][v]])
    tempdf.columns = names
    df0 = pd.concat([df0,tempdf],axis=1)

  mappedembeddings = pd.concat([mappedembeddings,df0],axis=0)
  
mappedembeddings.index = encodedindex