ml-enabler
ml-enabler copied to clipboard
Parametrize prediction keys
Prediction data is stored as JSONB. This means we can store any arbitrary key that's produced by the model. At the moment, ml-enabler API is built to aggregate two keys - ml_prediction
and osm_building_area
, because both the predictors are dealing with building data. The ml-enabler-cli, will produce the payload that can be readily submitted to the api.
In the future, when a new predictor is added, the cli can appropriately set the keys. At that time, the API would have to do one or more of the following:
- Accept keys as query params to aggregate
- Store keys that can be aggregated in the
Prediction
table -- so it doesn't confuse with things like URLs or other string values