Optimization Result Class

Stores multi-objective optimization results from NSGA-II/III including the Pareto-optimal solutions, feature pool, and training data. This class is returned by optimize_panel() and serves as input to fit_panel() for model fitting.

Slots

solutions

Data frame with one row per Pareto solution. Contains columns: solution_id (integer), base_features (list-column of character vectors with original feature names), features (list-column with transformed feature names used for model fitting), and one column per objective with numeric values.

feature_pool

Character vector of base features considered during optimization.

control

List of optimization control parameters including feature_transform.

training_signature

List with metadata about the training dataset.

aggregated_x

Raw (untransformed) training feature matrix containing base features. The feature transform is applied on-the-fly during fit_panel().

aggregated_y

Training response vector (factor).

aggregated_cohort

Cohort membership factor for training samples.

history

Optional per-generation NSGA history. Empty list when record_history = FALSE (default); otherwise a data frame with columns generation, individual, rank, is_pareto, and one column per objective (in user-facing direction). See nsga_history().