Interpolating a Time-Series Curve using Cubic Spline

Question

I have the following dataframe:

    M            P
2022-01-10      250
2022-04-03      100
2022-05-26      350
2022-08-01      200
   ...          ...

^just an example. Basically I have some dates and some prices associated with them.

I'm trying to create a daily time-series curve using the above fixed points with the cubic splining method.

I've looked at using the following -

from scipy.interpolate import CubicSpline

cs = CubicSpline(df['M'], df['P'])

But I get the following error -

TypeError: float() argument must be a string or a number, not 'datetime.date'

because it doesn't like the date variable type. Two questions:

a) Basically, how do I interpolate the x, y to produce my curve?

b) Secondly, I'm using this method to try and replicate the 'S' curve in the following documentation chart:

https://docs.scipy.org/doc/scipy/reference/generated/scipy.interpolate.CubicSpline.html

...however what I'm not them sure of is how I construct a daily time series with dates / calculated prices which corresponds to this curve? Will that automatically happen when the CubicSpline method is called?

Any help would be greatly appreciated!

Answer 1

Create DatetimeIndex by DataFrame.set_index , add missing values for daily frequency by DataFrame.asfreq and last use DataFrame.interpolate with method cubicspline :

'krogh', 'piecewise_polynomial', 'spline', 'pchip', 'akima', 'cubicspline': Wrappers around the SciPy interpolation methods of similar names. See Notes.

#convert date objects to datetimes
curve0['M'] = pd.to_datetime(curve0['M'])

df = curve0.set_index('M').asfreq('d').interpolate(method='cubicspline')
print (df)
                     P
M                     
2022-01-10  250.000000
2022-01-11  238.482096
2022-01-12  227.285579
2022-01-13  216.407317
2022-01-14  205.844180
               ...
2022-07-28  238.817308
2022-07-29  229.568679
2022-07-30  220.018338
2022-07-31  210.163156
2022-08-01  200.000000

[204 rows x 1 columns]