Skip to main content

Model.evaluate

Model.evaluate(targets, source, limit=None, metadata=None)

     This method evaluates a model given a source dataset for one or more targets

Parameters:​

       targets: Union[str, List[str], Dict[str, str]]
       One or more target names to evaluate on on.

       source: Union[Dataset, pd.DataFrame, str]
       Source dataset to evaluate on.

       limit: Optional[int]
       Number of records in the source to evaluate.

       metadata: Optional[bool]
       Optional boolean flag to return model name and version metadata in results.

Returns:​

       pd.DataFrame

Examples:​

Evaluate model on another Predibase dataset.

    dataset = pc.get_dataset("titanic_holdout_dataset")
    model.evaluate("Survived", dataset)

Evaluate model on a pandas dataframe.

    df = pd.read_csv("{path to file}/holdout_df.csv")
    model.evaluate("Survived", df)