algocomponents.tasks.evaluate_prediction

Submodules

algocomponents.tasks.evaluate_prediction.evaluate_prediction

class algocomponents.tasks.evaluate_prediction.evaluate_prediction.EvaluatePrediction(input_table: str, output_table: str, prediction_column: str, target_label_column: str, **kwargs)

Bases: SQLPipeline

A task to evaluate a prediction done by a model.

The task takes an input table with a prediction column and a target label column. The task calculates different evaluation metrics such as: true positives, false positives, true negatives, false negatives, accuracy, precision, recall, f1 score for different threshold boundaries.

Parameters:

• input_table – Table where model predictions and target labels exists.

• output_table – Table where results should be put.

• prediction_column – Column in input table for model predictions.

• target_label_column – Column in input table for model target label.

create_thresholds(n_thresholds) → str

Creates a query that produces a table with n_thresholds.

Parameters:

n_thresholds – How many thresholds, or how many rows, to produce.

Returns:

The query for creating n thresholds.

Contents