pandas.DataFrame.interpolate¶
- DataFrame.interpolate(method='linear', axis=0, limit=None, inplace=False, downcast='infer', **kwargs)¶
Interpolate values according to different methods.
Parameters : method : {‘linear’, ‘time’, ‘values’, ‘index’ ‘nearest’, ‘zero’,
‘slinear’, ‘quadratic’, ‘cubic’, ‘barycentric’, ‘krogh’, ‘polynomial’, ‘spline’ ‘piecewise_polynomial’, ‘pchip’}
- ‘linear’: ignore the index and treat the values as equally
spaced. default
- ‘time’: interpolation works on daily and higher resolution
data to interpolate given length of interval
‘index’: use the actual numerical values of the index
‘nearest’, ‘zero’, ‘slinear’, ‘quadratic’, ‘cubic’, ‘barycentric’, ‘polynomial’ is passed to scipy.interpolate.interp1d with the order given both ‘polynomial’ and ‘spline’ requre that you also specify and order (int) e.g. df.interpolate(method=’polynomial’, order=4)
- ‘krogh’, ‘piecewise_polynomial’, ‘spline’, and ‘pchip’ are all
wrappers around the scipy interpolation methods of similar names. See the scipy documentation for more on their behavior: http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html
axis : {0, 1}, default 0
- 0: fill column-by-column
- 1: fill row-by-row
limit : int, default None.
Maximum number of consecutive NaNs to fill.
inplace : bool, default False
Update the NDFrame in place if possible.
downcast : optional, ‘infer’ or None, defaults to ‘infer’
Downcast dtypes if possible.
Returns : Series or DataFrame of same shape interpolated at the NaNs
Examples
# Filling in NaNs: >>> s = pd.Series([0, 1, np.nan, 3]) >>> s.interpolate() 0 0 1 1 2 2 3 3 dtype: float64