Escalado y normalización de variables
Jeshua Romero Guadarrama- 3 Min para leer
- 05 May, 2021
Índice de contenido
Estandarización
Xescalada = (X - Xmedia)/Xstd
python
from sklearn import preprocessing
import numpy as np
import matplotlib.pyplot as plt
Con preprocessing.scale()
python
X_train = np.array([[ 1., -1., 2.],
[ 2., 0., 0.],
[ 0., 1., -1.]])
python
X_train.ndim
2
python
X_scaled = preprocessing.scale(X_train)
X_scaled
array([[ 0. , -1.22474487, 1.33630621],
[ 1.22474487, 0. , -0.26726124],
[-1.22474487, 1.22474487, -1.06904497]])
python
X_scaled.mean()
4.9343245538895844e-17
python
X_scaled.var()
1.0
Con preprocessing.StandardScaler()
python
scaler = preprocessing.StandardScaler()
python
fitted_scaler = scaler.fit(X_train)
fitted_scaler
StandardScaler()
python
fitted_scaler.mean_
array([1. , 0. , 0.33333333])
python
X_scaled_st = fitted_scaler.transform(X_train)
X_scaled_st
array([[ 0. , -1.22474487, 1.33630621],
[ 1.22474487, 0. , -0.26726124],
[-1.22474487, 1.22474487, -1.06904497]])
python
X_scaled_st.mean()
4.9343245538895844e-17
python
X_scaled_st.var()
1.0
Ejemplo para ver efecto de distorsión
python
X = [10000, 11000, 12000, 13000, 11000, 20000, 40000, 30000, 14000, 13000, 11000]
X
[10000, 11000, 12000, 13000, 11000, 20000, 40000, 30000, 14000, 13000, 11000]
python
plt.plot(X)
[<matplotlib.lines.Line2D at 0xb9517c0>]
python
X_scaled2 = preprocessing.scale(X)
X_scaled2
array([-0.741666 , -0.63288832, -0.52411064, -0.41533296, -0.63288832,
0.3461108 , 2.52166439, 1.4338876 , -0.30655528, -0.41533296,
-0.63288832])
python
plt.plot(X_scaled2)
[<matplotlib.lines.Line2D at 0xba55250>]
Escalado de datos en un rango
MinMaxScaler
python
from sklearn.preprocessing import MinMaxScaler
python
data = [[-1, 2], [-0.5, 6], [0, 10], [1, 18]]
data
[[-1, 2], [-0.5, 6], [0, 10], [1, 18]]
python
scaler = MinMaxScaler()
python
fitted_scaler = scaler.fit(data)
python
fitted_scaler.transform(data)
array([[0. , 0. ],
[0.25, 0.25],
[0.5 , 0.5 ],
[1. , 1. ]])
python
scaler_2 = MinMaxScaler(feature_range = (5,10))
python
fitted_scaler_2 = scaler_2.fit(data)
python
fitted_scaler_2.transform(data)
array([[ 5. , 5. ],
[ 6.25, 6.25],
[ 7.5 , 7.5 ],
[10. , 10. ]])
Ejemplo de distorsión
python
señal = np.array([50.2, 50., 48.2, 49., 53.1, 49.8, 49., 51.3])
señal
array([50.2, 50. , 48.2, 49. , 53.1, 49.8, 49. , 51.3])
python
plt.plot(señal)
plt.ylim(0,100)
(0.0, 100.0)
python
scaler_3 = MinMaxScaler()
fitted_scaler_3 = scaler_3.fit(señal.reshape(-1, 1))
señal_escalada = fitted_scaler_3.transform(señal.reshape(-1, 1))
señal_escalada
array([[0.40816327],
[0.36734694],
[0. ],
[0.16326531],
[1. ],
[0.32653061],
[0.16326531],
[0.63265306]])
python
plt.plot(señal_escalada)
[<matplotlib.lines.Line2D at 0xbcb6190>]
MaxAbsScaler
python
from sklearn.preprocessing import MaxAbsScaler
python
X = [[ 1., -1., 2.],
[ 2., 0., 0.],
[ 0., 1., -1.]]
python
transformer = MaxAbsScaler()
adj_transformer = transformer.fit(X)
results = adj_transformer.transform(X)
results
array([[ 0.5, -1. , 1. ],
[ 1. , 0. , 0. ],
[ 0. , 1. , -0.5]])
Normalización de vectores
python
from sklearn import preprocessing
python
X = [[ 1., -1., 2.],
[ 2., 0., 0.],
[ 0., 1., -1.]]
python
X_normalized_l1 = preprocessing.normalize(X, norm="l1")
X_normalized_l1
array([[ 0.25, -0.25, 0.5 ],
[ 1. , 0. , 0. ],
[ 0. , 0.5 , -0.5 ]])
python
X_normalized_l2 = preprocessing.normalize(X, norm="l2")
X_normalized_l2
array([[ 0.40824829, -0.40824829, 0.81649658],
[ 1. , 0. , 0. ],
[ 0. , 0.70710678, -0.70710678]])