Is there a faster wasy to create an exponential moving average than for loop?

Asked Nov 24 '20 at 13:35

Active Nov 24 '20 at 13:35

Viewed 194 times

Is there a faster way to create an exponential moving average than for loop?

I am priming the first n values of exp_aves to be mean of first n obseravtions.

def exp_mov_ave(observations, n):

    exp_ave = np.mean(observations[0:n])    
    
    lambdak = (n - 1) / (n + 1)
    exp_aves = []

    for i,element in enumerate(observations):
        if (i >= n):
            exp_ave = (lambdak * exp_ave) + ((1 - lambdak) * element)

        exp_aves.append(exp_ave)

    return np.array(exp_aves)

asked Nov 24 '20 at 13:35

ManInMoon

6,795
15
70
133

Did you try pandas function EWM ? – Johnson Francis Nov 24 '20 at 13:41
Is this one relevant? https://stackoverflow.com/questions/42869495/numpy-version-of-exponential-weighted-moving-average-equivalent-to-pandas-ewm – Jussi Nurminen Nov 24 '20 at 13:41
No. I understand pandas is slow – ManInMoon Nov 24 '20 at 13:42
@JussiNurminen That solution has a lot of code - seems overkill for a simple EMA – ManInMoon Nov 24 '20 at 13:44

Is there a faster wasy to create an exponential moving average than for loop?

0 Answers0