Analyze data
Access hub data
hubUtils (R) hubData (R) hubdata (Python) Accessing data (R) Accessing data (Python)
You can connect to any hub with the hubUtils and hubData R packages or the hubdata Python package, and be assured that all the data within the hub are validated against the hub data standards.
In R:
- connect to hub data with
hubData::connect_hub() - connect to a hub on S3 with
hubData::s3_bucket() - transform your existing data with
hubUtils::as_model_out_tbl()
In Python:
- connect to hub data with
hubdata.connect_hub() - read the model output as a PyArrow table with
HubConnection.to_table() - connect to target data with
hubdata.connect_target_data()
Get information about a hub
The hubUtils R package provides utilities that give you useful information about a hub:
- fetch a hub configuration file with
hubUtils::read_config() - get the submission timezone with
hubUtils::get_hub_timezone() - find out what the submission rounds are with
hubUtils::get_round_ids() - task ID names with
hubUtils::get_task_id_names()
In Python, hubdata.connect_hub() returns a HubConnection whose .admin and .tasks attributes expose the hub’s admin.json and tasks.json as dictionaries — giving you the submission rounds, task IDs, file formats, and other configuration.
Analyze hub data
You can use the hubEnsembles, hubEvals, and hubVis R packages to do any of the following:
- create a simple ensemble with
hubEnsembles::simple_ensemble() - create a linear pool ensemble with
hubEnsembles::linear_pool() - evaluate model performance with
hubEvals::score_model_out() - visualize model projections with
hubVis::plot_step_ahead_model()
Important packages
When analysing data from a hub, you will want the following packages:
- hubData seamlessly access data from a local or remote hub
- hubEnsembles combine models into an ensemble using built-in or custom methods
- hubVis visualize model output
- hubEvals compare and evaluate models in an established hub