Python Time Series ETS Decomposition

ts_ets_decomposition.py

# Assume Pandas DataFrame or Series

#moving average
df.ts_col.rolling(window=6).mean() #example 6 months rolling window
df.ts_col.rolling(window=12).mean() #example 12 months rolling window

#exponentially weighted moving average
df.ts_col.ewm(spane=12).mean() 

from statsmodels.tsa.holtwinters import SimpleExpSmoothing
span = 12
alpha = 2/(span+1)
df['SES12']=SimpleExpSmoothing(df['Thousands of Passengers']).fit(smoothing_level=alpha,optimized=False).fittedvalues.shift(-1)

#hodrick prescot filter
from statsmodels.tsa.filters.hp_filter import hpfilter
gdp_cycle, gdp_trend = hpfilter(df['realgdp'], lamb=1600)

#seasonal decomposition
#remember to freq because statsmodels uses it
df.index.freq = 'MS'
from statsmodels.tsa.seasonal import seasonal_decompose
result = seasonal_decompose(airline['Thousands of Passengers'], model='multiplicative')  # model='mul' also works
result.plot();

#double exponential smoothing
from statsmodels.tsa.holtwinters import ExponentialSmoothing
#additive trend
df['DESadd12'] = ExponentialSmoothing(df['Thousands of Passengers'], trend='add').fit().fittedvalues.shift(-1)
#multiplicative trend
df['DESmul12'] = ExponentialSmoothing(df['Thousands of Passengers'], trend='mul').fit().fittedvalues.shift(-1)

#triple exponential smoothing
# trend is linear and seasonality is additive
df['TESadd12'] = ExponentialSmoothing(df['Thousands of Passengers'],trend='add',seasonal='add',seasonal_periods=12).fit().fittedvalues
# trend is exponential and seasonality is multiplicative
df['TESmult12'] = ExponentialSmoothing(df['Thousands of Passengers'],trend='mul',seasonal='mul',seasonal_periods=12).fit().fittedvalues