So, reading BORIS files could be quite interesting! It would enable supervised classification of behaviours, e.g. in combination with {tidymodels}. They should not be imported to attempt matching the format movement format (xy), but rather into a form with start and end times. similar to their raw files (https://github.com/olivierfriard/BORIS/tree/master/tests/files). We could then have a join_behaviours or something similar, that then added the BORIS annotations to the movement data frames. Needs some careful thought though!
So, reading BORIS files could be quite interesting! It would enable supervised classification of behaviours, e.g. in combination with {tidymodels}. They should not be imported to attempt matching the format movement format (xy), but rather into a form with start and end times. similar to their raw files (https://github.com/olivierfriard/BORIS/tree/master/tests/files). We could then have a
join_behaviours
or something similar, that then added the BORIS annotations to the movement data frames. Needs some careful thought though!Note that Simba (https://www.nature.com/articles/s41593-024-01649-9) simply uses a random forest classifier, which is simple to implement with tidymodels. And it means we can easily change the model types without relying on other software! (see e.g. https://www.kirenz.com/blog/posts/2021-02-17-r-classification-tidymodels/#model-building which also uses XGBoost, Neural Net, KNN etc.
This is really promising actually - would bring behavioural classification to R for pretty much the first time I think!
For later development though, not v1 (maybe).