Optimization

Optimization task using machine learning models.

This task performs optimization on experimental data by:

1. Training multiple machine learning models (Random Forest, XGBoost, CatBoost)
2. Selecting the best performing model based on cross-validation R² scores
3. Using algorithms (NSGA-II or GA) to find optimal solutions
4. Generating comprehensive optimization results and analysis files

The optimization process considers:

• Target variables: Variables to maximize during optimization
• Constraints: Manual bounds on input features
• Thresholds: Minimum acceptable values for target variables

Generated Output Files:

• generalized_solutions.csv: All optimization solutions found
• best_generalized_solution.csv: Best solution based on CV and target values
• actual_vs_predicted.csv: Model validation data (observed vs predicted)
• feature_importance_matrix.csv: Feature importance for each target variable
• constraints_used_in_optimization.csv: Bounds applied to each feature
• optimization_progress.csv: Convergence history during optimization

Inputs: data (Table): Experimental data containing features and target variables targets_thresholds (JSONDict): Minimum threshold values for each target variable manual_constraints (JSONDict): Custom bounds for input features in format: {"feature_name": {"lower_bound": value, "upper_bound": value}}

Outputs: results_folder (Folder): Directory containing all optimization results and analysis files

Example: For a chemical process optimization, you might want to maximize yield and purity while keeping temperature below 100°C and pressure above 2 bar, with minimum yield of 80% and minimum purity of 95%.