Variables categóricas

Índice de contenido

Una variable categórica solo puede tener un número limitado, y generalmente fijo, de valores posibles (categorías). Algunos ejemplos de variables de este tipo pueden ser el género y el tipo de sangre. Pueden tener un orden (por ejemplo, ‘muy de acuerdo’ podría ir primero y ‘muy en desacuerdo’ podría ir en último lugar), pero las operaciones numéricas (sumas, restas, divisiones, …) no son posibles.

python

import pandas as pd

Convertir una columna en variable categórica

Convertir variable discreta

python

df = pd.read_csv("C:/Users/user/Desktop/adult-stretch.data", header=None, names = ["Color", "Size", "Other_column", "Age", "Target"])
df

	Color	Size	Other_column	Age	Target
0	YELLOW	SMALL	STRETCH	ADULT	T
1	YELLOW	SMALL	STRETCH	CHILD	T
2	YELLOW	SMALL	DIP	ADULT	T
3	YELLOW	SMALL	DIP	CHILD	F
4	YELLOW	SMALL	DIP	CHILD	F
5	YELLOW	LARGE	STRETCH	ADULT	T
6	YELLOW	LARGE	STRETCH	CHILD	T
7	YELLOW	LARGE	DIP	ADULT	T
8	YELLOW	LARGE	DIP	CHILD	F
9	YELLOW	LARGE	DIP	CHILD	F
10	PURPLE	SMALL	STRETCH	ADULT	T
11	PURPLE	SMALL	STRETCH	CHILD	T
12	PURPLE	SMALL	DIP	ADULT	T
13	PURPLE	SMALL	DIP	CHILD	F
14	PURPLE	SMALL	DIP	CHILD	F
15	PURPLE	LARGE	STRETCH	ADULT	T
16	PURPLE	LARGE	STRETCH	CHILD	T
17	PURPLE	LARGE	DIP	ADULT	T
18	PURPLE	LARGE	DIP	CHILD	F
19	PURPLE	LARGE	DIP	CHILD	F

python

df.columns

Index(['Color', 'Size', 'Other_column', 'Age', 'Target'], dtype='object')

python

df.dtypes

Color           object
Size            object
Other_column    object
Age             object
Target          object
dtype: object

python

df["Color"] = df["Color"].astype("category")
df.dtypes

Color           category
Size              object
Other_column      object
Age               object
Target            object
dtype: object

Convertir variable continua

python

df2 = pd.DataFrame({"Edad": np.random.randint(0, 100, 20)})
df2

	Edad
0	94
1	40
2	53
3	34
4	62
5	38
6	61
7	90
8	35
9	75
10	64
11	78
12	43
13	73
14	87
15	60
16	54
17	24
18	26
19	68

python

labels = ["{0} - {1}".format(i, i+9) for i in range(0,100,10)]
labels

['0 - 9',
 '10 - 19',
 '20 - 29',
 '30 - 39',
 '40 - 49',
 '50 - 59',
 '60 - 69',
 '70 - 79',
 '80 - 89',
 '90 - 99']

python

df2["Grupo_edad"] = pd.cut(df2.Edad, range(0,105,10), right = False, labels = labels)
df2

	Edad	Grupo_edad
0	94	90 - 99
1	40	40 - 49
2	53	50 - 59
3	34	30 - 39
4	62	60 - 69
5	38	30 - 39
6	61	60 - 69
7	90	90 - 99
8	35	30 - 39
9	75	70 - 79
10	64	60 - 69
11	78	70 - 79
12	43	40 - 49
13	73	70 - 79
14	87	80 - 89
15	60	60 - 69
16	54	50 - 59
17	24	20 - 29
18	26	20 - 29
19	68	60 - 69

Renombrar categorías

python

df.Color.unique()

['YELLOW', 'PURPLE']
Categories (2, object): ['YELLOW', 'PURPLE']

python

df.Color.cat.rename_categories({"YELLOW": "AMARILLO", "PURPLE": "MORADO"}, inplace=True)

python

df

	Color	Size	Other_column	Age	Target
0	AMARILLO	SMALL	STRETCH	ADULT	T
1	AMARILLO	SMALL	STRETCH	CHILD	T
2	AMARILLO	SMALL	DIP	ADULT	T
3	AMARILLO	SMALL	DIP	CHILD	F
4	AMARILLO	SMALL	DIP	CHILD	F
5	AMARILLO	LARGE	STRETCH	ADULT	T
6	AMARILLO	LARGE	STRETCH	CHILD	T
7	AMARILLO	LARGE	DIP	ADULT	T
8	AMARILLO	LARGE	DIP	CHILD	F
9	AMARILLO	LARGE	DIP	CHILD	F
10	MORADO	SMALL	STRETCH	ADULT	T
11	MORADO	SMALL	STRETCH	CHILD	T
12	MORADO	SMALL	DIP	ADULT	T
13	MORADO	SMALL	DIP	CHILD	F
14	MORADO	SMALL	DIP	CHILD	F
15	MORADO	LARGE	STRETCH	ADULT	T
16	MORADO	LARGE	STRETCH	CHILD	T
17	MORADO	LARGE	DIP	ADULT	T
18	MORADO	LARGE	DIP	CHILD	F
19	MORADO	LARGE	DIP	CHILD	F

Añadir categorías

python

df.Color.cat.add_categories("AZUL")

0     AMARILLO
1     AMARILLO
2     AMARILLO
3     AMARILLO
4     AMARILLO
5     AMARILLO
6     AMARILLO
7     AMARILLO
8     AMARILLO
9     AMARILLO
10      MORADO
11      MORADO
12      MORADO
13      MORADO
14      MORADO
15      MORADO
16      MORADO
17      MORADO
18      MORADO
19      MORADO
Name: Color, dtype: category
Categories (3, object): ['MORADO', 'AMARILLO', 'AZUL']

Eliminar categorías

python

df.Color.cat.remove_unused_categories()

0     AMARILLO
1     AMARILLO
2     AMARILLO
3     AMARILLO
4     AMARILLO
5     AMARILLO
6     AMARILLO
7     AMARILLO
8     AMARILLO
9     AMARILLO
10      MORADO
11      MORADO
12      MORADO
13      MORADO
14      MORADO
15      MORADO
16      MORADO
17      MORADO
18      MORADO
19      MORADO
Name: Color, dtype: category
Categories (2, object): ['MORADO', 'AMARILLO']

Ordenar por categoría

python

df.sort_values(by="Color", ascending=True)

	Color	Size	Other_column	Age	Target
19	MORADO	LARGE	DIP	CHILD	F
17	MORADO	LARGE	DIP	ADULT	T
16	MORADO	LARGE	STRETCH	CHILD	T
15	MORADO	LARGE	STRETCH	ADULT	T
14	MORADO	SMALL	DIP	CHILD	F
13	MORADO	SMALL	DIP	CHILD	F
12	MORADO	SMALL	DIP	ADULT	T
11	MORADO	SMALL	STRETCH	CHILD	T
10	MORADO	SMALL	STRETCH	ADULT	T
18	MORADO	LARGE	DIP	CHILD	F
0	AMARILLO	SMALL	STRETCH	ADULT	T
7	AMARILLO	LARGE	DIP	ADULT	T
6	AMARILLO	LARGE	STRETCH	CHILD	T
5	AMARILLO	LARGE	STRETCH	ADULT	T
4	AMARILLO	SMALL	DIP	CHILD	F
3	AMARILLO	SMALL	DIP	CHILD	F
2	AMARILLO	SMALL	DIP	ADULT	T
1	AMARILLO	SMALL	STRETCH	CHILD	T
8	AMARILLO	LARGE	DIP	CHILD	F
9	AMARILLO	LARGE	DIP	CHILD	F

python

df2.dtypes

Edad             int64
Grupo_edad    category
dtype: object

python

df2.sort_values(by="Grupo_edad", ascending=True)

	Edad	Grupo_edad
17	24	20 - 29
18	26	20 - 29
3	34	30 - 39
5	38	30 - 39
8	35	30 - 39
1	40	40 - 49
12	43	40 - 49
2	53	50 - 59
16	54	50 - 59
19	68	60 - 69
4	62	60 - 69
6	61	60 - 69
10	64	60 - 69
15	60	60 - 69
9	75	70 - 79
13	73	70 - 79
11	78	70 - 79
14	87	80 - 89
7	90	90 - 99
0	94	90 - 99

Agrupar por categoría

python

df2.groupby("Grupo_edad").size()

Grupo_edad
0 - 9      0
10 - 19    0
20 - 29    2
30 - 39    3
40 - 49    2
50 - 59    2
60 - 69    5
70 - 79    3
80 - 89    1
90 - 99    2
dtype: int64

Tratamiento de variables categóricas

Eliminar variables categóricas

La manera más simple de atajar el problema de las variables categóricas consiste en eliminarlas del conjunto de datos. Sin embargo, este enfoque puede provocar una pérdida de información potencialmente valiosa para el modelo de Machine Learning.

Método de label encoding

El método de label encoding (o codificación de etiqueta) consiste en asignar un número entero a cada valor diferente de una variable. El problema de este enfoque es que implica asumir un orden en las categorías, y las variables categóricas no siempre llevan implícito un orden claro. Por tanto, esta técnica es apropiada cuando se están tratando variables ordinales.

python

df

	Color	Size	Other_column	Age	Target
0	YELLOW	SMALL	STRETCH	ADULT	T
1	YELLOW	SMALL	STRETCH	CHILD	T
2	YELLOW	SMALL	DIP	ADULT	T
3	YELLOW	SMALL	DIP	CHILD	F
4	YELLOW	SMALL	DIP	CHILD	F
5	YELLOW	LARGE	STRETCH	ADULT	T
6	YELLOW	LARGE	STRETCH	CHILD	T
7	YELLOW	LARGE	DIP	ADULT	T
8	YELLOW	LARGE	DIP	CHILD	F
9	YELLOW	LARGE	DIP	CHILD	F
10	PURPLE	SMALL	STRETCH	ADULT	T
11	PURPLE	SMALL	STRETCH	CHILD	T
12	PURPLE	SMALL	DIP	ADULT	T
13	PURPLE	SMALL	DIP	CHILD	F
14	PURPLE	SMALL	DIP	CHILD	F
15	PURPLE	LARGE	STRETCH	ADULT	T
16	PURPLE	LARGE	STRETCH	CHILD	T
17	PURPLE	LARGE	DIP	ADULT	T
18	PURPLE	LARGE	DIP	CHILD	F
19	PURPLE	LARGE	DIP	CHILD	F

python

color_map = {"YELLOW": 1, "PURPLE": 2}
df["Color_numeric"] = df.Color.map(color_map)
df

	Color	Size	Other_column	Age	Target	Color_numeric
0	YELLOW	SMALL	STRETCH	ADULT	T	1
1	YELLOW	SMALL	STRETCH	CHILD	T	1
2	YELLOW	SMALL	DIP	ADULT	T	1
3	YELLOW	SMALL	DIP	CHILD	F	1
4	YELLOW	SMALL	DIP	CHILD	F	1
5	YELLOW	LARGE	STRETCH	ADULT	T	1
6	YELLOW	LARGE	STRETCH	CHILD	T	1
7	YELLOW	LARGE	DIP	ADULT	T	1
8	YELLOW	LARGE	DIP	CHILD	F	1
9	YELLOW	LARGE	DIP	CHILD	F	1
10	PURPLE	SMALL	STRETCH	ADULT	T	2
11	PURPLE	SMALL	STRETCH	CHILD	T	2
12	PURPLE	SMALL	DIP	ADULT	T	2
13	PURPLE	SMALL	DIP	CHILD	F	2
14	PURPLE	SMALL	DIP	CHILD	F	2
15	PURPLE	LARGE	STRETCH	ADULT	T	2
16	PURPLE	LARGE	STRETCH	CHILD	T	2
17	PURPLE	LARGE	DIP	ADULT	T	2
18	PURPLE	LARGE	DIP	CHILD	F	2
19	PURPLE	LARGE	DIP	CHILD	F	2

python

df.dtypes

Color            category
Size               object
Other_column       object
Age                object
Target             object
Color_numeric    category
dtype: object

python

df.Color.cat.codes

0     1
1     1
2     1
3     1
4     1
5     1
6     1
7     1
8     1
9     1
10    0
11    0
12    0
13    0
14    0
15    0
16    0
17    0
18    0
19    0
dtype: int8

python

df["Color_codes"] = df.Color.cat.codes
df

	Color	Size	Other_column	Age	Target	Color_numeric	Color_codes
0	YELLOW	SMALL	STRETCH	ADULT	T	1	1
1	YELLOW	SMALL	STRETCH	CHILD	T	1	1
2	YELLOW	SMALL	DIP	ADULT	T	1	1
3	YELLOW	SMALL	DIP	CHILD	F	1	1
4	YELLOW	SMALL	DIP	CHILD	F	1	1
5	YELLOW	LARGE	STRETCH	ADULT	T	1	1
6	YELLOW	LARGE	STRETCH	CHILD	T	1	1
7	YELLOW	LARGE	DIP	ADULT	T	1	1
8	YELLOW	LARGE	DIP	CHILD	F	1	1
9	YELLOW	LARGE	DIP	CHILD	F	1	1
10	PURPLE	SMALL	STRETCH	ADULT	T	2	0
11	PURPLE	SMALL	STRETCH	CHILD	T	2	0
12	PURPLE	SMALL	DIP	ADULT	T	2	0
13	PURPLE	SMALL	DIP	CHILD	F	2	0
14	PURPLE	SMALL	DIP	CHILD	F	2	0
15	PURPLE	LARGE	STRETCH	ADULT	T	2	0
16	PURPLE	LARGE	STRETCH	CHILD	T	2	0
17	PURPLE	LARGE	DIP	ADULT	T	2	0
18	PURPLE	LARGE	DIP	CHILD	F	2	0
19	PURPLE	LARGE	DIP	CHILD	F	2	0

python

df.Color.cat.ordered

False

Método de one-hot encoding

El método de one-hot encoding (o codificación one-hot) consiste en crear nuevas columnas indicando la presencia o ausencia de cada posible valor de una variable categórica.

One-hot encoding, al contrario que el método de label encoding, no implica asumir un orden en las categorías. Por tanto, esta técnica es apropiada para tratar variables nominales. Sin embargo, es importante tener en cuenta que el método de one-hot encoding no suele dar buenos resultados si hay muchos valores diferentes para la variable categórica que se está procesando (generalmente, más de 15).

python

pd.get_dummies(data=df, columns=["Color"], drop_first=True)

	Size	Other_column	Age	Target	Color_numeric	Color_codes	Color_YELLOW
0	SMALL	STRETCH	ADULT	T	1	1	1
1	SMALL	STRETCH	CHILD	T	1	1	1
2	SMALL	DIP	ADULT	T	1	1	1
3	SMALL	DIP	CHILD	F	1	1	1
4	SMALL	DIP	CHILD	F	1	1	1
5	LARGE	STRETCH	ADULT	T	1	1	1
6	LARGE	STRETCH	CHILD	T	1	1	1
7	LARGE	DIP	ADULT	T	1	1	1
8	LARGE	DIP	CHILD	F	1	1	1
9	LARGE	DIP	CHILD	F	1	1	1
10	SMALL	STRETCH	ADULT	T	2	0	0
11	SMALL	STRETCH	CHILD	T	2	0	0
12	SMALL	DIP	ADULT	T	2	0	0
13	SMALL	DIP	CHILD	F	2	0	0
14	SMALL	DIP	CHILD	F	2	0	0
15	LARGE	STRETCH	ADULT	T	2	0	0
16	LARGE	STRETCH	CHILD	T	2	0	0
17	LARGE	DIP	ADULT	T	2	0	0
18	LARGE	DIP	CHILD	F	2	0	0
19	LARGE	DIP	CHILD	F	2	0	0

python

df.dtypes

Color            category
Size               object
Other_column       object
Age                object
Target             object
Color_numeric    category
Color_codes          int8
dtype: object

python

pd.get_dummies(data=df)

	Color_codes	Color_PURPLE	Color_YELLOW	Size_LARGE	Size_SMALL	Other_column_DIP	Other_column_STRETCH	Age_ADULT	Age_CHILD	Target_F	Target_T	Color_numeric_2	Color_numeric_1
0	1	0	1	0	1	0	1	1	0	0	1	0	1
1	1	0	1	0	1	0	1	0	1	0	1	0	1
2	1	0	1	0	1	1	0	1	0	0	1	0	1
3	1	0	1	0	1	1	0	0	1	1	0	0	1
4	1	0	1	0	1	1	0	0	1	1	0	0	1
5	1	0	1	1	0	0	1	1	0	0	1	0	1
6	1	0	1	1	0	0	1	0	1	0	1	0	1
7	1	0	1	1	0	1	0	1	0	0	1	0	1
8	1	0	1	1	0	1	0	0	1	1	0	0	1
9	1	0	1	1	0	1	0	0	1	1	0	0	1
10	0	1	0	0	1	0	1	1	0	0	1	1	0
11	0	1	0	0	1	0	1	0	1	0	1	1	0
12	0	1	0	0	1	1	0	1	0	0	1	1	0
13	0	1	0	0	1	1	0	0	1	1	0	1	0
14	0	1	0	0	1	1	0	0	1	1	0	1	0
15	0	1	0	1	0	0	1	1	0	0	1	1	0
16	0	1	0	1	0	0	1	0	1	0	1	1	0
17	0	1	0	1	0	1	0	1	0	0	1	1	0
18	0	1	0	1	0	1	0	0	1	1	0	1	0
19	0	1	0	1	0	1	0	0	1	1	0	1	0

python

# USO DE SKLEARN

	Edad
0	94
1	40
2	53
3	34
4	62
5	38
6	61
7	90
8	35
9	75
10	64
11	78
12	43
13	73
14	87
15	60
16	54
17	24
18	26
19	68

	Color_codes	Color_PURPLE	Color_YELLOW	Size_LARGE	Size_SMALL	Other_column_DIP	Other_column_STRETCH	Age_ADULT	Age_CHILD	Target_F	Target_T	Color_numeric_2	Color_numeric_1
0	1	0	1	0	1	0	1	1	0	0	1	0	1
1	1	0	1	0	1	0	1	0	1	0	1	0	1
2	1	0	1	0	1	1	0	1	0	0	1	0	1
3	1	0	1	0	1	1	0	0	1	1	0	0	1
4	1	0	1	0	1	1	0	0	1	1	0	0	1
5	1	0	1	1	0	0	1	1	0	0	1	0	1
6	1	0	1	1	0	0	1	0	1	0	1	0	1
7	1	0	1	1	0	1	0	1	0	0	1	0	1
8	1	0	1	1	0	1	0	0	1	1	0	0	1
9	1	0	1	1	0	1	0	0	1	1	0	0	1
10	0	1	0	0	1	0	1	1	0	0	1	1	0
11	0	1	0	0	1	0	1	0	1	0	1	1	0
12	0	1	0	0	1	1	0	1	0	0	1	1	0
13	0	1	0	0	1	1	0	0	1	1	0	1	0
14	0	1	0	0	1	1	0	0	1	1	0	1	0
15	0	1	0	1	0	0	1	1	0	0	1	1	0
16	0	1	0	1	0	0	1	0	1	0	1	1	0
17	0	1	0	1	0	1	0	1	0	0	1	1	0
18	0	1	0	1	0	1	0	0	1	1	0	1	0
19	0	1	0	1	0	1	0	0	1	1	0	1	0

	Edad
0	94
1	40
2	53
3	34
4	62
5	38
6	61
7	90
8	35
9	75
10	64
11	78
12	43
13	73
14	87
15	60
16	54
17	24
18	26
19	68

	Color_codes	Color_PURPLE	Color_YELLOW	Size_LARGE	Size_SMALL	Other_column_DIP	Other_column_STRETCH	Age_ADULT	Age_CHILD	Target_F	Target_T	Color_numeric_2	Color_numeric_1
0	1	0	1	0	1	0	1	1	0	0	1	0	1
1	1	0	1	0	1	0	1	0	1	0	1	0	1
2	1	0	1	0	1	1	0	1	0	0	1	0	1
3	1	0	1	0	1	1	0	0	1	1	0	0	1
4	1	0	1	0	1	1	0	0	1	1	0	0	1
5	1	0	1	1	0	0	1	1	0	0	1	0	1
6	1	0	1	1	0	0	1	0	1	0	1	0	1
7	1	0	1	1	0	1	0	1	0	0	1	0	1
8	1	0	1	1	0	1	0	0	1	1	0	0	1
9	1	0	1	1	0	1	0	0	1	1	0	0	1
10	0	1	0	0	1	0	1	1	0	0	1	1	0
11	0	1	0	0	1	0	1	0	1	0	1	1	0
12	0	1	0	0	1	1	0	1	0	0	1	1	0
13	0	1	0	0	1	1	0	0	1	1	0	1	0
14	0	1	0	0	1	1	0	0	1	1	0	1	0
15	0	1	0	1	0	0	1	1	0	0	1	1	0
16	0	1	0	1	0	0	1	0	1	0	1	1	0
17	0	1	0	1	0	1	0	1	0	0	1	1	0
18	0	1	0	1	0	1	0	0	1	1	0	1	0
19	0	1	0	1	0	1	0	0	1	1	0	1	0

	Edad
0	94
1	40
2	53
3	34
4	62
5	38
6	61
7	90
8	35
9	75
10	64
11	78
12	43
13	73
14	87
15	60
16	54
17	24
18	26
19	68

	Color_codes	Color_PURPLE	Color_YELLOW	Size_LARGE	Size_SMALL	Other_column_DIP	Other_column_STRETCH	Age_ADULT	Age_CHILD	Target_F	Target_T	Color_numeric_2	Color_numeric_1
0	1	0	1	0	1	0	1	1	0	0	1	0	1
1	1	0	1	0	1	0	1	0	1	0	1	0	1
2	1	0	1	0	1	1	0	1	0	0	1	0	1
3	1	0	1	0	1	1	0	0	1	1	0	0	1
4	1	0	1	0	1	1	0	0	1	1	0	0	1
5	1	0	1	1	0	0	1	1	0	0	1	0	1
6	1	0	1	1	0	0	1	0	1	0	1	0	1
7	1	0	1	1	0	1	0	1	0	0	1	0	1
8	1	0	1	1	0	1	0	0	1	1	0	0	1
9	1	0	1	1	0	1	0	0	1	1	0	0	1
10	0	1	0	0	1	0	1	1	0	0	1	1	0
11	0	1	0	0	1	0	1	0	1	0	1	1	0
12	0	1	0	0	1	1	0	1	0	0	1	1	0
13	0	1	0	0	1	1	0	0	1	1	0	1	0
14	0	1	0	0	1	1	0	0	1	1	0	1	0
15	0	1	0	1	0	0	1	1	0	0	1	1	0
16	0	1	0	1	0	0	1	0	1	0	1	1	0
17	0	1	0	1	0	1	0	1	0	0	1	1	0
18	0	1	0	1	0	1	0	0	1	1	0	1	0
19	0	1	0	1	0	1	0	0	1	1	0	1	0