熊猫中的数据框列的索引

Question

The main problem is to create a list full of categorical factor's indices

There is dataframe with lots of columns types of which were detemined before importing file with pd.read_csv() .

dtypes = {
    ...
    'Format_type': 'category',
    'Geo_new': 'category',
    'Age_min': 'int16',
    'Age_max': 'int16',
    'Sex': 'category',
    ...}

So I made a table with columns names and their indices , and than take categorical colums by myself

col_list = [i for i in (df.columns.get_values())]
idx_list = [i for i in range(len(df.columns.get_values()))]
column_num = pd.DataFrame(data = {'column_name': col_list,
                                  'idx_list': idx_list})
column_num

Than get table of columns name column_name and indices idx_list

column_name idx_list
...
Format_type 5
Geo_new     6
Age_min     7
Age_max     8
Sex         9
...

and insert categorical columns indices in the list:

categorical_features = [...5, 6, 9...]

Thus, i fill list by myself. Is there the way to create list of columns, which values are calegory automatically?

Answer 1

I believe you need DataFrame.select_dtypes with Index.get_indexer for indices:

df = pd.DataFrame({
        'A':list('abcdef'),
         'B':pd.Categorical([4,5,4,5,5,4]),
         'C':[7,8,9,4,2,3],
         'D': pd.Categorical([1,3,5,7,1,0]),
         'E':[5,3,6,9,2,4],
         'F':list('aaabbb')
})

c = df.select_dtypes('category').columns
print (c)
Index(['B', 'D'], dtype='object')

i = df.columns.get_indexer(df.select_dtypes('category').columns)
print (i)
[1 3]

Also your code should be simlify:

col_list = df.columns.tolist()
idx_list = range(len(col_list))
column_num = pd.DataFrame(data = {'column_name': col_list, 'idx_list': idx_list})

Answer 2

还有另一种方法！

categorical_list = list(np.where(df.dtypes == 'category')[0])