[Machine Learning] Data preprocessing: convert categorical data into numerical values

 1 import pandas as pd
 2 from io import StringIO
 3 
 4 csv_data = '''A,B,C,D
 5 1,2,3,4
 6 5,6,,8
 7 0,11,12,'''
 8 
 9 df = pd.read_csv(StringIO(csv_data))
10 print(df)
11 #统计为空的数目
12 print(df.isnull().sum())
13 print(df.values)
14 
15 #丢弃空的
16 print(df.dropna())
17 print('after', df)
18 from sklearn.preprocessing import Imputer
19 # axis=0 列   axis = 1 行
20 imr = Imputer(missing_values='NaN', strategy='mean', axis=0)
21 imr.fit(df) # fit  构建得到数据
22 imputed_data = imr.transform(df.values) #transform 将数据进行填充
23 print(imputed_data)
24 
25 df = pd.DataFrame([['green', 'M', 10.1, 'class1'],
26                    ['red', 'L', 13.5, 'class2'],
27                    ['blue', 'XL', 15.3, 'class1']])
28 df.columns =['color', 'size', 'price', 'classlabel']
29 print(df)
30 
31 size_mapping = {'XL':3, 'L':2, 'M':1}
32 df['size'] = df['size'].map(size_mapping)
33 print(df)
34 
35 ## 遍历Series
36 for idx, label in enumerate(df['classlabel']):
37     print(idx, label)
38 
39 #1, 利用LabelEncoder类快速编码,但此时对color并不适合,
40 #看起来,好像是有大小的
41 from sklearn.preprocessing import LabelEncoder
42 class_le = LabelEncoder()
43 color_le = LabelEncoder()
44 df['classlabel'] = class_le.fit_transform(df['classlabel'].values)
45 #df['color'] = color_le.fit_transform(df['color'].values)
46 print(df)
47 
48 #2, 映射字典将类标转换为整数
49 import numpy as np
50 class_mapping = {label: idx for idx, label in enumerate(np.unique(df['classlabel']))}
51 df['classlabel'] = df['classlabel'].map(class_mapping)
52 print('2,', df)
53 
54 
55 #3,处理1不适用的
56 #利用创建一个新的虚拟特征
57 from sklearn.preprocessing import OneHotEncoder
58 pf = pd.get_dummies(df[['color']])
59 df = pd.concat([df, pf], axis=1)
60 df.drop(['color'], axis=1, inplace=True)
61 print(df)