Optimization of hyperparameters in CALS systems using the ant colony method using the example of forecasting the volumes of air passenger and cargo transportation

In the modern world, civil aviation is actively developing in the direction of import substitution and the transition to the use of domestic aircraft, which requires the implementation of CALS technologies, including analytical and simulation models. These models depend on many parameters, which makes the task of optimizing hyperparameters relevant. This work is devoted to the construction of SARIMA models for forecasting passenger and cargo transportation in the Russian Federation. For optimal forecasting of 25 aggregated monthly parameters, separate SARIMA models were developed, and the purpose of the study is to apply modifications of the ant colony method to determine the optimal values of hyperparameters. The results show the effectiveness of the ant colony method for optimizing the hyperparameters of SARIMA and LSTM models. Optimization was carried out using the MAE and RMSE metrics, with less than 10% of solutions from the Pareto set corresponding to the optimal values. However, for more complex models, the search time for optimal solutions increases significantly, without providing a significant advantage over simple SARIMA models. The practical significance of the work lies in determining the optimal values ​​of hyperparameters for 25 indicators of passenger and cargo transportation, as well as in forecasting the needs for the aircraft fleet until 2030. The projected growth in demand requires an additional expansion of the aircraft fleet by 30 units per year, which, taking into account the decommissioning, implies the production of 60 new aircraft of all types.