Forecasting Average Land Temperature using FBProphet
The data contain date in avarage land temperature in a month.
Data plot
- auto prophet : using prophet default parameters values
- auto multiplicative mode : using prophet default parameter values but change seasonality_mode to multiplicative
- all_seasonality : add all seasonality (only try with as previous example in workshop)
- yearly_seasonality : as we plot we know the data has seasonality yearly
- yearly_seasonality_with_add : using seasonality yearly and add seasonality_mode to additive
models = {"auto_prophet":Prophet(),
"auto_multiplicative_mode":Prophet(seasonality_mode = 'multiplicative'),
"all_seasonality" : Prophet(growth= 'linear',
seasonality_mode = 'multiplicative',
daily_seasonality= False,
weekly_seasonality= False,
yearly_seasonality= False).add_seasonality(
name='monthly',
period=30.5,
fourier_order=15,
prior_scale=15
).add_seasonality(
name='weekly',
period=7,
fourier_order=10,
prior_scale=20
).add_seasonality(
name='yearly',
period=365.25,
fourier_order=20
).add_seasonality(
name='quarterly',
period=365.25/4,
fourier_order=5,
prior_scale=15
),
"yearly_seasonality" : Prophet(growth= 'linear',
seasonality_mode = 'multiplicative',
daily_seasonality= False,
weekly_seasonality= False,
yearly_seasonality= False).add_seasonality(
name='yearly',
period=365,
fourier_order=20
),
"yearly_seasonality_with_add" : Prophet(growth= 'linear',
seasonality_mode = 'additive',
daily_seasonality= False,
weekly_seasonality= False,
yearly_seasonality= False).add_seasonality(
name='yearly',
period=365,
fourier_order=20
)
}
From table above we know that the best model is from prophet models that use seasonality yearly and seasonality mode additive. It's suitable with analysis data above that the seasonality when we plot is yearly and if we observe to the graph we can see thet the seasonality more fitted as additive seasonality rather than multiplicative seasonality. I got excellent explenantion for the difference between additive and multiplicative seasonality mode.
source : https://medium.com/@sigmundojr/seasonality-in-python-additive-or-multiplicative-model-d4b9cf1f48a7
