TypeError:float()参数必须是字符串或数字,而不是'Period'

Question

我有一个像这样的列的pandas数据框:

df.columns = pd.to_datetime(list(df)) #list(df) = ["2017-01", "2016-01", ...]

然后我在数据集的每一行中执行插值,因为我有一些我想要摆脱的NaN.这是打印的结果:

ORIGINAL  
2007-12-01     NaN 
2008-12-01     NaN 
2009-12-01     NaN 
2010-12-01   -0.35 
2011-12-01    0.67 
2012-12-01     NaN 
2013-12-01     NaN 
2014-12-01    1.03 
2015-12-01    0.37 
2016-12-01     NaN 
2017-12-01     NaN 
Name: row1, dtype: float64 

INTERPOLATION  
2007-12-01   -0.350000 
2008-12-01   -0.350000 
2009-12-01   -0.350000 
2010-12-01   -0.350000 
2011-12-01    0.670000 
2012-12-01    0.790219 
2013-12-01    0.910109 
2014-12-01    1.030000 
2015-12-01    0.370000 
2016-12-01    0.370000 
2017-12-01    0.370000 
Name: row1, dtype: float64

然后我尝试绘制插值行并得到:

TypeError: float() argument must be a string or a number, not 'Period' 

整个代码:

print("ORIGINAL\n", series)
interpolation = series.interpolate(method=func, limit=10, limit_direction='both')
interpolation.plot()
print("INTERPOLATION\n",interpolation)

在我看来,错误是在系列中的时间值,但我认为matplotlib应该很容易处理它,所以我肯定做错了.提前致谢.

Answer 1

这是最简单的答案，无需升级或降级熊猫。

pd.plotting.register_matplotlib_converters()

有时注册会导致另一个错误，例如compute.use_bottleneck，use_numexpr错误，该错误会消除该调用注销：P

喜欢： pd.plotting.deregister_matplotlib_converters()

来源：链接

Answer 2

如果我这样做，它将起作用：

plt.plot(row.index, row.values)
plt.show()

我不知道为什么

Answer 3

复制您的插值结果

df = pd.read_clipboard(header=None)
df.columns = ['Period','Value']
df['Period'] = pd.to_datetime(df['Period'])
df  = df.set_index('Period')
print(df)

               Value
Period              
2007-12-01 -0.350000
2008-12-01 -0.350000
2009-12-01 -0.350000
2010-12-01 -0.350000
2011-12-01  0.670000
2012-12-01  0.790219
2013-12-01  0.910109
2014-12-01  1.030000
2015-12-01  0.370000
2016-12-01  0.370000
2017-12-01  0.370000


df.plot()