ctf4science.eval_module.long_time_forecast_dynamical

ctf4science.eval_module.long_time_forecast_dynamical(truth: ndarray, prediction: ndarray, modes: int, bins: int) → float

Compute the long-time forecast score for dynamical systems (histogram-based).

Uses the last modes time steps per feature, builds L1-normalized histograms with bins bins, and compares truth vs prediction. Score is 100 * (1 - relative_L1_error) so that 100 is perfect.

Parameters:

truth(T, F) ndarray

Ground truth, shape (time steps, features).

prediction(T, F) ndarray

Predicted data, same shape as truth.

modesint

Number of last time steps to use per feature.

binsint

Number of histogram bins.

Returns:

float

Long-time forecast score as a percentage (100 = perfect).