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