No abstract

No abstract

Gabion weirs are environment-friendly structures widely used for irrigation and drainage network purposes. These structures' hydraulic performance is fundamentally different from solid weirs' due to their porosity and the existence of a through-flow discharge. This paper investigates the reliability and suitability of a number of Machine learning models for estimation of hydraulic performance of gabion weirs.  Generally, three different Boosting ensemble models, including Gradient Boosting, XGBoost, and CatBoost, are compared to the well-known Random Forest and a Stacked Regression model, with respect to their accuracy in prediction of the discharge coefficient and through-flow discharge ratio of gabion weirs in free flow conditions. The Bayesian optimization approach is used to fine-tune model hyper-parameters automatically. Recursive feature elimination analysis is also performed to find optimum combination of features for each model. Results indicate that the CatBoost model has outperformed other models in terms of estimating the through flow discharge ratio (Qin/Qt) with R2 = 0.982, while both XGBoost and CatBoost models have shown close performance in terms of estimating the discharge coefficient (Cd) with R2 of CatBoost equal to 0.994 and R2 of XGBoost equal to 0.992. Weakest results were also produced by Decision tree regressor with R2 = 0.821 and 0.865 for estimation of Cd and Qin/Qt values.

Predicting the behaviour of wildfires can help save lives and reduce health, socioeconomic, and environmental impacts. Because wildfire behaviour is highly dependent on fuel type and distribution, their accurate estimation is paramount for accurate prediction of the fire propagation dynamics. This paper studies the effect of combining automated hyperparameter tuning with Bayesian optimisation and recursive feature elimination on the accuracy of three boosting (AdaB, XGB, CatB), two bagging (Random Forest, Extremely Randomised Trees), and three stacking ensemble models with respect to their ability to estimate the vegetation cover type from cartographic data. The models are trained on the University of California Irvine (UCI) cover type dataset using five-fold cross-validation. Feature importance scores are calculated and used in recursive feature elimination analysis to study the sensitivity of model accuracy to the different feature combinations. Our results indicate that the implemented fine-tuning procedure significantly affects the accuracy of all models investigated, with XGB achieving an overall accuracy of 97.1% slightly outperforming the others.