All Ramp analyses require a DataContext.
A DataContext represents the environment of the
analysis. Most importantly this means for a given store
and pandas index value, Ramp will consider the data immutable –
it will not check the data again to see if it has changed.
Args
- store: An instance of store.Store or a path. If a path
- Ramp will default to an HDFPickleStore at that path
if PyTables is installed, a PickleStore otherwise.
Defaults to MemoryStore.
- data: a pandas DataFrame. If all data has been precomputed this
- may not be required.
- train_index: a pandas Index specifying the data instances to be
- used in training. Stored results will be cached against this.
If not provided, the entire index of data will be used.
- prep_index: a pandas Index specifying the data instances to be
- used in prepping (“x” values). Stored results will be cached against this.
If not provided, the entire index of data will be used.
-
load_context(name)
Loads a previously saved context with given name,
assigning the stored training and prep indices
and returning any stored config.
-
save_context(name, config=None)
Saves this context (specifically it’s train and prep indices)
to it’s store with the given
name, along with the config, if provided.