I can use this code to calculate an exponential moving average in Pandas:
df['ewm_40'] = df['response'].ewm(span=40, adjust=True).mean()
When there are some volatile data points, there can be a sharp increase in the exponential moving average that is undesirable.
How would I cap the exponential moving average so no point moves more than 15% from the prior point? I want to balance stability while still reflecting new data.
IIUC, I compute the exponential moving average and update the values in order if they are increasing/decreasing more than a certain percentage.
df = pd.DataFrame({'test': [101, 103, 170, 10, 113]})
df['ewm_40'] = df['test'].ewm(span=40, adjust=True).mean()
print(df.ewm_40)
mx = 0.15
def sharp_increase(df, mx):
return any(abs(df.ewm_40 / df.ewm_40.shift() - 1) > mx + 0.00001)
while sharp_increase(df, mx):
perc = df.ewm_40 / df.ewm_40.shift()
tf = list(abs(perc - 1) > mx)
index = tf.index(True)
if perc[2] > 1 + mx:
multiplier = 1 + mx
else:
multiplier = 1 - mx
df.loc[index, 'ewm_40'] = df.loc[index - 1, 'ewm_40'] * multiplier
print(df.ewm_40)
Prints:
0 101.000000
1 102.025000
2 125.825453
3 94.662094
4 98.705363
Name: ewm_40, dtype: float64
0 101.000000
1 102.025000
2 117.328750
3 94.662094
4 98.705363
Name: ewm_40, dtype: float64
0 101.000000
1 102.025000
2 117.328750
3 99.729437
4 98.705363
Name: ewm_40, dtype: float64