GroupBy 索引和列然后转换 Pandas 中的选定列

Question

我有一个示例 DF：

sample_df_train = pd.DataFrame(np.random.randint(1,20,size=(10, 3)), columns=list('ABC'))
sample_df_train["date"]= ["2020-02-01","2020-02-01","2020-02-01","2020-02-01","2020-02-01",
                "2020-02-02","2020-02-02","2020-02-02","2020-02-02","2020-02-02"]
sample_df_train["date"] = pd.to_datetime(sample_df_train["date"])
sample_df_train.set_index(sample_df_train["date"],inplace=True)
del sample_df_train["date"]
sample_df_train["A_cat"] = ["ind","sa","sa","sa","ind","ind","sa","sa","ind","sa"]
sample_df_train["B_cat"] = ["sa","ind","ind","sa","sa","sa","ind","sa","ind","sa"]
sample_df_train

操作：

            A   B   C   A_cat   B_cat
  date                  
2020-02-01  13  13  14  ind      sa
2020-02-01  3   2   10  sa       ind
2020-02-01  2   6   6   sa       ind
2020-02-01  11  6   8   sa       sa
2020-02-01  4   9   1   ind      sa
2020-02-02  3   3   18  ind      sa
2020-02-02  17  3   17  sa       ind
2020-02-02  1   5   17  sa       sa
2020-02-02  13  15  9   ind      ind
2020-02-02  12  16  19  sa       sa

我正在尝试根据 2 个条件转换此 DF：

1. GroupBy based on Index and some columns
2. Transform (add 100 to column value) selected columns in the DF based on the GroupBy op.

知识产权：

group_by_cols = ['date', 'A_cat']  Be noted that "date' is index. 
selected_columns = ["A"]

代码：

sample_df_train[selected_columns] = sample_df_train.reset_index().groupby(group_by_cols)[selected_columns].apply(lambda x: x+100)

操作：

             A  B   C   A_cat   B_cat
 date                   
2020-02-01  NaN 17  11  ind     sa
2020-02-01  NaN 10  9   sa      ind
2020-02-01  NaN 2   11  sa      ind
2020-02-01  NaN 3   16  sa      sa
2020-02-01  NaN 7   3   ind     sa
2020-02-02  NaN 6   5   ind     sa
2020-02-02  NaN 19  3   sa      ind
2020-02-02  NaN 4   15  sa      sa
2020-02-02  NaN 11  8   ind     ind
2020-02-02  NaN 14  14  sa      sa

预期操作：

The entire DF with 100 added to values in Column A.

我不明白为什么我会得到NANs 。 任何建议都会很棒。

Answer 1

请注意，您在执行groupby时只是reset_index ，因此原始 df 的index与groupby结果不同

sample_df_train= sample_df_train.reset_index()
sample_df_train[selected_columns] = sample_df_train.groupby(group_by_cols)[selected_columns].apply(lambda x: x+100)