从另一个数据框架的选定列中填充NaN值

Question

我有这样的df1

id        name  level personality      type  weakness    atk    def     hp  stage
0    53.0     Persian   40.0        mild    normal  fighting  104.0  116.0    NaN    2.0
1   126.0      Magmar   44.0      docile       NaN     water   96.0   83.0  153.0    1.0
2    57.0    Primeape    9.0      lonely  fighting    flying    NaN   66.0   43.0    2.0
3     3.0    Venusaur   44.0       sassy     grass      fire  136.0  195.0   92.0    3.0
4    11.0     Metapod    4.0       naive     grass      fire    NaN  114.0    NaN    2.0
5   126.0      Magmar   96.0      modest      fire     water   62.0  114.0    NaN    1.0
6   137.0     Porygon   96.0     relaxed       NaN  fighting   68.0   50.0  127.0    1.0
7    69.0  Bellsprout   84.0      lonely     grass      fire    NaN    NaN    NaN    1.0
8    10.0    Caterpie    3.0     serious       NaN    flying    NaN    NaN   15.0    1.0
9    12.0  Butterfree   12.0       hasty       NaN    flying   20.0    NaN    NaN    3.0
10   35.0    Clefairy   18.0      impish     fairy    poison   33.0    NaN    NaN    1.0
11   59.0    Arcanine   35.0      gentle      fire     water   45.0   60.0   80.0    2.0
12  111.0     Rhyhorn   31.0     naughty      rock     water   40.0    NaN  175.0    1.0
13  136.0     Flareon   75.0        bold       NaN     water    NaN  143.0    NaN    2.0
14   51.0     Dugtrio   82.0      gentle    ground     water  152.0  161.0  168.0    2.0
15   38.0   Ninetales    5.0       brave      fire     water    NaN  179.0  173.0    2.0
16  102.0   Exeggcute   88.0        rash       NaN      fire    NaN  124.0    NaN    1.0 
........

and df2 as

weakness      type  count
3       fire     grass     11
10     water      fire      9
0   fighting    normal      6
4     flying  fighting      3
8     poison     fairy      3
6      grass     water      1
9       rock      fire      1
7     ground  electric      1

我想在类型列中使用df2更新NaN值，并在两个dfs中匹配弱点列。例如，在df1的第8行和第9行中，"type"为NaN赋值。我想用df2更新它们匹配df1中的弱点列。所以那些8,9类型的值应该是" fighting "等等。这有点像df2和df1之间的一对多关系。

我试着

df1.update(df2)

和

df1.fillna(df2)

但是他们没有给出期望的输出。如有任何帮助，不胜感激。

Answer 1

您可以从df2创建一个字典，弱点列作为键，类型列作为它们各自的值，然后使用该字典fillnadf1中的类型列使用map:

m = dict(zip(df2.weakness,df2.type))
df1.type = df1.type.fillna(df1.weakness.map(m))

打印:

>>> df1[['weakness','type']]
weakness      type
0   fighting    normal
1      water      fire
2     flying  fighting
3       fire     grass
4       fire     grass
5      water      fire
6   fighting    normal
7       fire     grass
8     flying  fighting
9     flying  fighting
10    poison     fairy
11     water      fire
12     water      rock
13     water      fire
14     water    ground
15     water      fire
16      fire     grass

Answer 2

1. 从df2创建一个系列，将weakness值映射到type值:

   mapping = df2.set_index("weakness")["type"]

2. 使用此映射映射df1["weakness"]来创建默认值:

   defaults = df1["weakness"].map(mapping)

3. 使用默认值作为fillna方法的参数:

   df1["type"] = df1["type"].fillna(defaults)

Answer 3

代码内联

# Merge both dataframes using "weakness" as key
df = pd.merge(df1, df2[['weakness', 'type']], 
on="weakness",  suffixes=("", "_y"), how="left")
# Replace nans
df['type'].fillna(df['type_y'], inplace=True)
# Drop additional columns resulted from Merge
df.drop(columns=['type_y'])