kinoml.analysis.metrics

Module Contents

kinoml.analysis.metrics.root_mean_squared_error(*args, **kwargs)

Returns the square-root of scikit-learn’s mean_squared_error metric. All arguments are forwarded to that function.

kinoml.analysis.metrics.performance(predicted, observed, verbose=True, n_boot=100, confidence=0.95, sample_ratio=0.8, _seed=1234)

Measure the predicted vs observed performance with different metrics (R2, MSE, MAE, RMSE).

Parameters

• predicted (array-like) – Data points predicted by the model.

• observed (array-like) – Observed data points, as available in the dataset.

• verbose (bool, optional=True) – Whether to print results to stdout.

• n_boot (int, optional=100) – Number of bootstrap iterations. Set to 1 to disable bootstrapping.

• confidence (float, optional=0.95) – Confidence interval, relative to 1. Default is 95%.

• sample_ratio (float, optional=0.8) – Proportion of data to sample in each iteration.

• _seed (int, optional=1234) – Random seed. Each bootstrap iteration gets a different seed based on this initial one.

Returns

results – This dictionary contains one item per metric (see above), with a 4-element tuple each: mean, standard deviation, and lower and upper bounds for the confidence interval.

Return type

dict of tuple

Note

TODO: Reimplement samples with scipy.stats.norm or with numpy.