AbstractScorer 类

所有计分器的基类。

继承
AbstractScorer

构造函数

AbstractScorer(metric_name: str, task: str, *args: Any, **kwargs: Any)

参数

metric_name
task

方法

calculate_lift

重写此方法以进行提升计算。

应在不同列上运行的所有不同转换之间进行一次同类比较。 输出应类似于相对分数增加或相对误差改进

is_experiment_better_than_baseline

重写以提供两个试验输出之间的比较。

is_experiment_score_better

如果 experiment_score 比 baseline_score_with_mod 好,则返回 true。

score

重写此方法以进行分数计算。

计算估计器的性能。

calculate_lift

重写此方法以进行提升计算。

应在不同列上运行的所有不同转换之间进行一次同类比较。 输出应类似于相对分数增加或相对误差改进

abstract calculate_lift(baseline_score: float, experiment_score: float) -> float

参数

baseline_score
必需
experiment_score
必需

is_experiment_better_than_baseline

重写以提供两个试验输出之间的比较。

abstract is_experiment_better_than_baseline(baseline_score: float, experiment_score: float, epsilon: float, **kwargs: Any) -> bool

参数

baseline_score
必需

基线分数。

experiment_score
必需

试验分数。

epsilon
必需

被视为增益/损失的最小增量。

kwargs
必需

与特定评分器相关的关键字参数。

返回

实验分数是否比基线更好。

is_experiment_score_better

如果 experiment_score 比 baseline_score_with_mod 好,则返回 true。

is_experiment_score_better(baseline_score_with_mod: float, experiment_score: float) -> bool

参数

baseline_score_with_mod
必需

带有 stat sig mod 的 baseline_score,例如 baseline_score + epsilon

experiment_score
必需

我们要试验的转换评分,用于确定是否有改进。

score

重写此方法以进行分数计算。

计算估计器的性能。

abstract score(estimator: sklearn.base.BaseEstimator, valid_features: numpy.ndarray, y_actual: numpy.ndarray) -> float

参数

estimator
必需
valid_features
必需
y_actual
必需