时间序列预处理之缺失值填充

发表于 更新于
时间序列缺失值的处理方法。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
import pandas as pd
import copy

ts = pd.read_csv('daily-min-temperatures.csv', parse_dates=['Date'], index_col='Date')
print(ts.head())

def ts_fillna(data, date_col, val_col):
    data = copy.deepcopy(data)
    # 构造辅助列
    helper = pd.DataFrame({date_col: pd.date_range(data.index.min(), data.index.max())})
    # 原数据与辅助列合并
    newdata = pd.merge(data, helper, on=date_col, how='outer').sort_values(date_col)

    # 固定值填充
    # newdata[val_col] = newdata[val_col].fillna(0)
    # 前向填充
    # newdata[val_col] = newdata[val_col].fillna(method='ffill')
    # 后向填充
    # newdata[val_col] = newdata[val_col].fillna(method='bfill')
    # 插值填充
    newdata[val_col] = newdata[val_col].interpolate(method='linear')

    # 索引重建
    newdata[date_col] = pd.to_datetime(newdata[date_col])
    newdata.set_index(date_col, inplace=True, verify_integrity=False)
    return newdata

ts2 = ts_fillna(ts, 'Date', 'Temp')
print(ts2.head())