3rd ICAI 2024

International Conference on Automotive Industry 2024

Mladá Boleslav, Czech Republic

tree within the ensemble. Rather than seeking the optimal split across all features, Extra Trees opts for random selection of feature subsets and thresholds. This strategy serves to mitigate overfitting and enhances the diversity among the trees. • Bootstrap Aggregating (Bagging): Like Random Forests, the Extra Trees Regressor also employs a technique known as bootstrap aggregating, or bagging. This method entails training each decision tree on a bootstrapped sample of the training data, which involves sampling the training dataset with replacement. This additional step bolsters the model’s resilience and diminishes variance. • Efficiency: The term “Extra” in Extra Trees denotes the heightened level of randomness introduced in contrast to Random Forests. Through the random selection of features and thresholds without the pursuit of optimal choices, Extra Trees exhibit potential computational efficiency, particularly beneficial for datasets with high dimensions. • Robustness to Noisy Data: Due to its randomization, the Extra Trees Regressor typically displays reduced sensitivity to noisy data and outliers when juxtaposed with other regression models. Its versatility extends to effectively managing various data types and distributions across a broad spectrum. • Tuning Parameters: Like any machine learning algorithm, the Extra Trees Regressor offers parameters that can be adjusted to enhance its performance. These parameters include the number of trees in the ensemble, the maximum depth of each tree, and the number of features to evaluate at each split. Tuning these parameters allows for the optimization of the model’s effectiveness for specific tasks.

Figure 3: The simplified process of random forest algorithm

Source: (Wei, 2022)

84