I have hurst exponent code.
below code, default lags_count value is 100.
but, when I change 'lags_count' value, the hurst exponent result also change.
and if time series data term is shorter than default 100, below function doesn't work.
so, I tried to lags_count value = time_series.shape[0]*0.7, 0.6, 0.5, etc.
I don't know what value is correct.
plus, I don't want to negative value of hurst exponent.
I think there is accurate or trust-worthy lags parameter value!
def calcHurstExponent(ts,lags):
tau = [np.sqrt(np.std(np.subtract(ts[lag:], ts[:-lag]))) for lag in lags]
poly = np.polyfit(np.log(lags), np.log(tau), 1)
result = poly[0]*2.0
return result
def get_hurst_exponent(df,target_column,lags_count=100):
lags = range(2, lags_count)
ts = np.log(df[target_column])
return calcHurstExponent(ts,lags) 