3rd ICAI 2024

International Conference on Automotive Industry 2024

Mladá Boleslav, Czech Republic

implementation and parallelization empower it to seamlessly manage datasets comprising millions of samples and thousands of features, underscoring its versatility and suitability for diverse real-world applications. In summary, the CatBoost Regressor emerges as a potent and adaptable algorithm for regression tasks, particularly excelling with datasets featuring a combination of categorical and numerical features. Its direct handling of categorical data, coupled with its robustness, efficiency, and scalability, positions it as a favored option across a spectrum of regression challenges spanning diverse domains.

Figure 4: The simplified process of random forest algorithm

Source: (Yang et al., 2023)

2.2.3 Selection of model evaluation metrics Selecting appropriate effectiveness metrics for predictive models is not a straightforward task. The complexity of the models and the multiplicity of problem classes have not made it particularly simple to select the appropriate effectiveness metrics for predictive models. In the relevant literature, there are a limited number of review articles related in content to the issue of the effectiveness of models in short-term electricity demand planning. In the study by Andrew, it may be concluded that MAPE and RMSE are the most used measures for prediction effectiveness. In the referred work, MAPE was

86