-1

How can I use python to impute timeseries data with seasonality components?

Below is an example of how my data looks like, I am missing data for long periods that includes many cycles and not sure how to solve that.

enter image description here

mohamed elhafiz
  • 69
  • 1
  • 8

2 Answers2

1

What you can do, is to have a look at Prophet - a Facebook package for time-series analysis. It does not impute data only, but it handles missing-data.

It breaks your data up in three functions; trend, seasonality and "holidays" ("holidays" you can treat as external feature e.g holidays, or days where you have campings, pay-check etc).

CutePoison
  • 4,679
  • 5
  • 28
  • 63
1

The library FEDOT offers a powerful solution for this. Check the notebook:

https://github.com/nccr-itmo/FEDOT/blob/master/notebooks/latest/5_ts_specific_cases.ipynb

Snapshot from the example in the notebook:

enter image description here

mohamed elhafiz
  • 69
  • 1
  • 8
  • The link is obsolete, the actual one is https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/5_ts_specific_cases.ipynb – Nikitin Nikolay Sep 17 '21 at 15:36