Refactoring fclib
hongooi73 opened this issue · 6 comments
hongooi73 commented
The current fclib modules are a legacy from the tsperf iteration of this repo. They have a number of issues:
- Contain functions which are idiosyncratic to the energy forecasting use case, eg computing lags for specific seasonality definitions
- Other functions carry implicit assumptions about the nature of the data, eg exactly 2 key/grain variables
- Lack of unit tests
These make fclib less suitable as a general-purpose toolkit for enabling forecasting. This issue is to track progress in refactoring the library into something that better fits this description.
hongooi73 commented
Should remove integer output arg for modelling utility funcs, this just adds needless complexity.
vapaunic commented
@hongooi73 , is this issue addressed with your latest work on code refactoring and unit tests?
hongooi73 commented
Not yet, there are more changes I want to make as we discussed before.
hongooi73 commented
- Refactor the one-hot-encoding functions to always return a fixed number of columns
vapaunic commented
Hi @hongooi73 , with the latest changes in one-hot-encoding functions, is this issue resolved?
hongooi73 commented
Yeah let's close this for now.