Convert timeseries data as Stationary

Hi,

I am getting below results for Dickey-Fuller Test. Can someone suggest technique to make my data stationary.

I could see that graph for actual data, rolling mean, rolling std is also almost same.

I have already tried log transformation, differencing and the results are still same.

Results of Dickey-Fuller Test:
Test Statistic -8.161630e+00
p-value 9.110753e-13
#Lags Used 2.000000e+01
Number of Observations Used 1.351000e+03
Critical Value (1%) -3.435200e+00
Critical Value (5%) -2.863682e+00
Critical Value (10%) -2.567910e+00
dtype: float64

Regards,
Sai

© Copyright 2013-2019 Analytics Vidhya