我使用numpy创建了一个给定大小的随机矩阵。对于时间序列模拟,我为相应的矩阵创建了一个频率为一个月的时间序列。现在,我想将它们组合起来,并将它们作为pandas数据帧。这就是我目前所拥有的
import numpy as np
import pandas as pd
cols = ['time', 'cases', 'deaths', 'recoveries']
data = np.random.randint(0,50,(50,3))
times = pd.date_range('2019-12-01', periods=50, freq='MS')
df = pd.DataFrame(pd.concat(times, data, ignore_index=True), columns=cols)
这在第8行给出了以下错误-
TypeError: cannot concatenate object of type '<class 'pandas._libs.tslibs.timestamps.Timestamp'>'; only Series and DataFrame objs are valid
因此,我尝试使用times = pd.Series(pd.date_range('2019-12-01', periods=50, freq='MS'))
将其转换为系列,但这反过来又产生了错误-
TypeError: first argument must be an iterable of pandas objects, you passed an object of type "Series"
预期O/p-
| time |cases|deaths|recoveries|
|------------------------------------|
| 2019-12-01 | 0 | 0 | 0 |
| 2020-01-01 | 1 | 0 | 0 |
| 2020-02-01 | 2 | 1 | 0 |
我建议创建DatetimeIndex
列,以便通过pandas的类似日期时间的方法进行处理:
#removed time column
cols = ['cases', 'deaths', 'recoveries']
data = np.random.randint(0,50,(50,3))
#added time in name parameter
times = pd.date_range('2019-12-01', periods=50, freq='MS', name='time')
#removed concat and added index parameter
df = pd.DataFrame(data, columns=cols, index=times)
print (df.head(10))
cases deaths recoveries
time
2019-12-01 28 44 25
2020-01-01 21 23 26
2020-02-01 15 17 5
2020-03-01 35 3 42
2020-04-01 46 7 3
2020-05-01 23 47 28
2020-06-01 31 30 34
2020-07-01 8 4 15
2020-08-01 46 14 24
2020-09-01 43 47 6
如果需要列,只添加DataFrame.reset_index
:
df = pd.DataFrame(data, columns=cols, index=times).reset_index()
print (df.head(10))
time cases deaths recoveries
0 2019-12-01 2 26 43
1 2020-01-01 43 40 41
2 2020-02-01 23 12 22
3 2020-03-01 43 37 28
4 2020-04-01 7 26 20
5 2020-05-01 19 46 41
6 2020-06-01 43 1 0
7 2020-07-01 19 42 4
8 2020-08-01 14 39 40
9 2020-09-01 15 8 25