skill_score.py

Calculates Skill Score.

SkillScore

chronax.skill_score.SkillScore · inherits BaseAggregator

Class to compute skill score for models compared to a baseline.

Skill score S_j quantifies how much model j reduces forecasting error compared to the fixed baseline model β on average.

__init__(self, pivot_table, baseline_model="seasonal_naive")

Initialize the SkillScore aggregator.

Parameter	Type	Default	Description
pivot_table	pd.DataFrame	-	DataFrame with models as index, tasks as columns, scores as values
baseline_model	str	"seasonal_naive"	Name of the baseline model for skill score calculation

__call__(self) -> pd.Series

Compute skill score for each model compared to baseline.

Formula: S_j = 1 - (Π_r clip(E_rj / E_rβ; l, u))^(1/R)

Where: - E_rj = error of model j on task r - E_rβ = error of baseline model β on task r - clip(x; l, u) = max(l, min(x, u)) - l = 10^-2 = 0.01 (lower bound) - u = 100 (upper bound) - R = number of tasks

Returns: pd.Series (Series with skill score for each model).