Custom Loss Functions in XGBoost Algorithm for Enhanced Critical Error Mitigation in Drill-Wear Analysis of Melamine-Faced Chipboard DOI Creative Commons
Michał Bukowski, Jarosław Kurek, Bartosz Świderski

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(4), P. 1092 - 1092

Published: Feb. 7, 2024

The advancement of machine learning in industrial applications has necessitated the development tailored solutions to address specific challenges, particularly multi-class classification tasks. This study delves into customization loss functions within eXtreme Gradient Boosting (XGBoost) algorithm, which is a critical step enhancing algorithm’s performance for applications. Our research motivated by need precision and efficiency domain, where implications misclassification can be substantial. We focus on drill-wear analysis melamine-faced chipboard, common material furniture production, demonstrate impact custom functions. paper explores several variants Weighted Softmax Loss Functions, including Edge Penalty Adaptive Loss, challenges class imbalance heightened importance accurately classifying edge classes. findings reveal that these significantly reduce errors without compromising overall accuracy model. not only contributes field providing nuanced approach function but also underscores context-specific adaptations algorithms. results showcase potential balancing efficiency, ensuring reliable effective settings.

Language: Английский

Cascading hazards from two recent glacial lake outburst floods in the Nyainqêntanglha range, Tibetan Plateau DOI Open Access
Menger Peng, Xue Wang, Guoqing Zhang

et al.

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 626, P. 130155 - 130155

Published: Sept. 14, 2023

Language: Английский

Citations

16
Applications of artificial intelligence technologies in water environments: From basic techniques to novel tiny machine learning systems DOI Creative Commons
Majid Bagheri,

Nakisa Farshforoush,

Karim Bagheri

et al.

Process Safety and Environmental Protection, Journal Year: 2023, Volume and Issue: 180, P. 10 - 22

Published: Sept. 30, 2023

Language: Английский

Citations

16
Weighted multi-error information entropy based you only look once network for underwater object detection DOI
Haiping Ma, Yajing Zhang, Shengyi Sun

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2023, Volume and Issue: 130, P. 107766 - 107766

Published: Dec. 27, 2023

Language: Английский

Citations

15
Comparative Analysis of Artificial Intelligence Methods for Streamflow Forecasting DOI Creative Commons
Wei Yaxing, Huzaifa Hashim, Sai Hin Lai

et al.

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 10865 - 10885

Published: Jan. 1, 2024

Deep learning excels at managing spatial and temporal time series with variable patterns for streamflow forecasting, but traditional machine algorithms may struggle complicated data, including non-linear multidimensional complexity. Empirical heterogeneity within watersheds limitations inherent to each estimation methodology pose challenges in effectively measuring appraising hydrological statistical frameworks of variables. This study emphasizes forecasting the region Johor, a coastal state Peninsular Malaysia, utilizing 28-year streamflow-pattern dataset from Malaysia's Department Irrigation Drainage Johor River its tropical rainforest environment. For this dataset, wavelet transformation significantly improves resolution lag noise when historical data are used as lagged input variables, producing 6% reduction root-mean-square error. A comparative analysis convolutional neural networks artificial reveals these models' distinct behavioral patterns. Convolutional exhibit lower stochasticity than dealing complex transformed into format suitable modeling. However, suffer overfitting, particularly cases which structure is overly simplified. Using Bayesian networks, we modeled network weights biases probability distributions assess aleatoric epistemic variability, employing Markov chain Monte Carlo bootstrap resampling techniques. modeling allowed us quantify uncertainty, providing confidence intervals metrics robust quantitative assessment model prediction variability.

Language: Английский

Citations

5
Custom Loss Functions in XGBoost Algorithm for Enhanced Critical Error Mitigation in Drill-Wear Analysis of Melamine-Faced Chipboard DOI Creative Commons
Michał Bukowski, Jarosław Kurek, Bartosz Świderski

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(4), P. 1092 - 1092

Published: Feb. 7, 2024

The advancement of machine learning in industrial applications has necessitated the development tailored solutions to address specific challenges, particularly multi-class classification tasks. This study delves into customization loss functions within eXtreme Gradient Boosting (XGBoost) algorithm, which is a critical step enhancing algorithm’s performance for applications. Our research motivated by need precision and efficiency domain, where implications misclassification can be substantial. We focus on drill-wear analysis melamine-faced chipboard, common material furniture production, demonstrate impact custom functions. paper explores several variants Weighted Softmax Loss Functions, including Edge Penalty Adaptive Loss, challenges class imbalance heightened importance accurately classifying edge classes. findings reveal that these significantly reduce errors without compromising overall accuracy model. not only contributes field providing nuanced approach function but also underscores context-specific adaptations algorithms. results showcase potential balancing efficiency, ensuring reliable effective settings.

Language: Английский

Citations

5