Usage

You can run springtime as a command-line tool in a terminal or use it as a python library e.g. in a Jupyter notebook. Below, we explain both CLI and API.

CLI to run recipes

The main component of springtime command-line is the recipe (scientific workflow). A recipe is a file with yaml extension that includes a set of instructions to reproduce a certain result. Recipes are written in a nice and readable format, e.g.

datasets:
  npn_obs:
    dataset: RNPN
    species_ids:
      functional_type: "Deciduous broadleaf" # multiple species
    phenophase_ids:
        name: breaking leaf buds
    years: [2015, 2020]
    area:
      name: Washington
      bbox:
        [
          -124.08406940413612,
          45.50277198520317,
          -117.39620059586387,
          49.99938001479683,
        ]
  daymet:
    dataset: daymet_multiple_points
    points:
      source: npn_obs
    years: [2015, 2020]
    variables:
      - tmin
      - tmax
    resample:
      frequency: month
      operator: median
dropna: True
experiment:
  experiment_type: regression  # --> pycaret.regression.RegressionExperiment
  setup:
    ... # setup of the experiment
  init_kwargs:
    ... # intial arguments for models
  compare_models:
    include:
      - 'lr'  # linear regression
      - 'rf'  # random forest regressor
      - 'sklearn.svm.SVR'
      - 'interpret.glassbox.ExplainableBoostingRegressor'
    cross_validation: true
    n_select: 2
  plots:
    - error
    - residuals

Such a recipe can then be executed with springtime command in a terminal:

springtime model_comparison_usecase.yaml

We provide several "recipes" for running experiments.

Python API

Springtime is written in Python (with parts in R) and can also be used in an interactive (IPython/Jupyter) session. For example:

Downloading data

from springtime.datasets.PEP725Phenor import PEP725Phenor
dataset = PEP725Phenor(species='Syringa vulgaris')
dataset.download()
df = dataset.load()

We provide several notebooks for downloading data from various sources. See "Datasets" documentation.