mae#
- tslearn.metrics.performance.mae(y_true, y_pred, ts_weights=None, timestamps_weights=None, multioutput='uniform_average')[source]#
Mean absolute error (MAE)
MAE is a measure of the prediction accuracy for forecasting and regression tasks that computes the average of the deviations between ground truth and predicted values.
- Parameters:
- y_true: array like, shape (n_ts, sz, d)
Target dataset of ground_truth values
- y_pred: array like, shape (n_ts, sz, d)
Estimated dataset of predicted values
- ts_weights: array like, shape (n_ts,) or None (default: None)
Weights to apply to the time-series in the datasets if non-uniform. Use none for uniform weights.
- timestamps_weights: array like, shape (sz,) or None (default: None)
Weights to apply to the timestamps of each-time series if non-uniform. Use none for uniform weights.
- multioutput: {‘uniform_average’, ‘raw_values’} or array-like, shape (d,) (default: ‘uniform_average’)
for multivariate timeseries, defines the aggregation of per feature results, if any.
- ‘raw_values’:
no aggregation, result is per feature
- ‘uniform_average’:
errors of all features are averaged with uniform weights
- array-like:
errors of all features are averaged using the given weights
- Returns:
- float or array-like, shape (d,)
If multioutput is ‘raw_values’, then mean absolute error is returned for each feature. Otherwise, the average of each feature is returned.
