search

kdgutier / esrnn_torch

MIT License
156 stars 44 forks source link

How to set seasonality #25

Closed ramdhan1989 closed 3 years ago

ramdhan1989 commented 3 years ago

Hi, by default seasonality is set to []. my data is daily data and have multiple seasonality : weekly and every six months. how to put in the model ? would you mind explaining the help function below seasonality: int list list of seasonalities of the time series Hourly [24, 168], Daily [7], Weekly [52], Monthly [12], Quarterly [4], Yearly []

why for daily sesonality is 7 , weekly 52, etc ?

thanks

kdgutier commented 3 years ago

DAILY A very common seasonality for daily data is 7 because there are common patterns that repeat every week. Particularly if your data comes from human economic activities. Of course there are 'natural' seasonalities for daily data like 365, the length of the year, but usually you want to train your model with shorter input_sizes.

HOURLY The list allows you to add multiple seasonalities at once for example with hourly data you may want both the daily seasonality and the weekly seasonality so you can set seasonality = [24, 24*7].

WEEKLY Weekly is set to 52 because there are 52 weeks in the year.

Regarding your data if you have data recorded every six months, your seasonality can be 2. At the end of the day the model you select should follow empirical performance.

ramdhan1989 commented 3 years ago

Ok got it

thanks a lot