dlpy.metrics.roc_auc_score¶

dlpy.metrics.roc_auc_score(y_true, y_score, pos_label, castable=None, cutstep=0.001, id_vars=None)¶

Compute the area under the receiver operating characteristic (ROC) curve for binary classification tasks.

Parameters:

y_true : string or CASColumn: The column of the ground truth labels. If it is a string, then y_score has to be a string and they both belongs to the same CASTable specified by the castable argument. If it is a CASColumn, then y_score has to be a CASColumn, and the castable argument is ignored. When both y_score and y_true are CASColumn, they can be in different CASTable.
y_score : string or CASColumn: The column of estimated probability for the positive class. If it is a string, then y_true has to be a string and they both belongs to the same CASTable specified by the castable argument. If it is a CASColumn, then y_true has to be a CASColumn, and the castable argument is ignored. When both y_score and y_true are CASColumn, they can be in different CASTable.
pos_label : string, int or float: The positive class label.
castable : CASTable, optional: The CASTable object to use as the source if the y_score and y_true are strings. Default = None
cutstep : float > 0 and < 1, optional: The stepsize of threshold cutoffs. Default=0.001.
id_vars : string or list of strings, optional: Column names that serve as unique id for y_true and y_score if they are from different CASTables. The column names need to appear in both CASTables, and they serve to match y_true and y_score appropriately, since observation orders can be shuffled in distributed computing environment. Default = None.

Returns: