Train an ELUC predictor

[ofrancon]

Train a model that predicts ELUC give land use changes

Once #17 is done, the dataset contains data like:
ELUC, primf_diff, primn_diff, secdf_diff, secdn_diff, urban_diff, c3ann_diff, ...

The diff columns are the land use change, i.e. the features
ELUC is the emissions from land use change, i.e. the label

Maybe the current state is also needed to improve the predictions: primf, primn, secdf, secdn, urban, .... They could be additional features.

Evaluation of the predictions can be done on different test sets. The split could be for instance:

• Train / Test split by years: test on last 15 years
• Train / Test split by geographical regions
• Train / Test random

[ofrancon]

Additional features could be:

• land usage and diffs from previous years (time series)
• data from the neighboring cells (geographical cluster)

[ofrancon]

The link to the aggregated dataset can be found in #14

[ofrancon]

We have plenty of ELUC predictors now: see folder use_cases/eluc/predictors