特征工程之特征预处理_0

特征工程系列数据链接：https://pan.baidu.com/s/1ZUwOM206B-YUzaNzebi_cg 提取码:7w86
from sklearn.preprocessing import StandardScaler, MinMaxScaler, RobustScaler
import numpy as np
import pandas as pd
np.set_printoptions(suppress=True)

简单小数据

views = pd.DataFrame([1295., 25., 19000., 5., 1., 300.], columns=['views'])
views

	views
0	1295.0
1	25.0
2	19000.0
3	5.0
4	1.0
5	300.0

Standard Scaler $\frac{x_i-\mu }{\sigma }$

ss = StandardScaler()
views['zscore'] = ss.fit_transform(views[['views']])
views

	views	zscore
0	1295.0	-0.307214
1	25.0	-0.489306
2	19000.0	2.231317
3	5.0	-0.492173
4	1.0	-0.492747
5	300.0	-0.449877

vw = np.array(views['views'])
(vw[0] - np.mean(vw)) / np.std(vw)

-0.30721413311687235

Min-Max Scaler $\frac{x_i-min(x)}{max(x)-min(x)}$

mms = MinMaxScaler()
views['minmax'] = mms.fit_transform(views[['views']])
views

	views	zscore	minmax
0	1295.0	-0.307214	0.068109
1	25.0	-0.489306	0.001263
2	19000.0	2.231317	1.000000
3	5.0	-0.492173	0.000211
4	1.0	-0.492747	0.000000
5	300.0	-0.449877	0.015738

(vw[0] - np.min(vw)) / (np.max(vw) - np.min(vw))

0.068108847834096528

Robust Scaler $\frac{x_i-median(x)}{IQ{R}_{(1,3)}(x)}$

rs = RobustScaler()
views['robust'] = rs.fit_transform(views[['views']])
views

	views	zscore	minmax	robust
0	1295.0	-0.307214	0.068109	1.092883
1	25.0	-0.489306	0.001263	-0.132690
2	19000.0	2.231317	1.000000	18.178528
3	5.0	-0.492173	0.000211	-0.151990
4	1.0	-0.492747	0.000000	-0.155850
5	300.0	-0.449877	0.015738	0.132690

quartiles = np.percentile(vw, (25., 75.))
iqr = quartiles[1] - quartiles[0]
(vw[0] - np.median(vw)) / iqr

1.0928829915560916