缺失值

Python 预测分析进阶

Nele Verbiest

Senior Data Scientist @PythonPredictions

用聚合值替换缺失值 (1)

donor_id	age
5	-
3	25
2	36
8	40
1	26

用聚合值替换缺失值 (2)

donor_id	age
5	38
3	25
2	36
8	40
1	26

平均年龄：38

用聚合值替换缺失值 (3)

donor_id	max_donation
5	-
3	1 000 000
2	100
8	40
1	120

max_donation 的均值：25 065

max_donation 的中位数：110

用聚合值替换缺失值 (4)

donor_id	max_donation
5	110
3	1 000 000
2	100
8	40
1	120

max_donation 的均值：25 065

max_donation 的中位数：110

用固定值替换缺失值 (1)

donor_id	sum_donations
5	130
3	10
2	-
8	40
1	120

用固定值替换缺失值 (2)

donor_id	sum_donations
5	130
3	10
2	0
8	40
1	120

在 Python 中替换缺失值

# Replace missing values by 0
replacement = 0
basetable["donations_last_year"] = 
    basetable["donations_last_year"].fillna(replacement)

# Replace missing values by mean
replacement = basetable["age"].mean()
basetable["age"] = basetable["age"].fillna(replacement)

缺失值哑变量

    donor_id email
0     32770  [email protected]
1     32776  nan
2     32777  [email protected]
3     65552  nan

basetable["no_email"] = pd.Series(
                            [0 if email==email else 1 
                            for email in basetable["email"]])

      donor_id email                    no_email
0     32770  [email protected]   0
1     32776  nan                        1
2     32777  [email protected]   0
3     65552  nan                        1

Passons à la pratique !

Python 预测分析进阶