Performance Evaluation of Hybrid PSO-BPNN-AdaBoost and PSO-BPNN-XGBoost Models for Rockburst Prediction with Imbalanced Datasets DOI Creative Commons
Shujian Li, Pengpeng Lu, Weizhang Liang

и другие.

Applied Sciences, Год журнала: 2024, Номер 14(24), С. 11792 - 11792

Опубликована: Дек. 17, 2024

The rockburst hazard is a primary geological disaster endangering the environment in underground engineering. Due to complexity of mechanism, traditional methods are insufficient predict objectively, especially when dealing with an imbalanced dataset. To address this issue, hybrid models PSO-BPNN-AdaBoost and PSO-BPNN-XGBoost were developed hazards study. First, dataset 266 cases was constructed, containing six indicators: maximum tangential stress, uniaxial compressive strength, tensile elastic deformation energy index, stress brittleness coefficient strength. Then, original oversampled using synthetic minority oversampling technique (SMOTE) for balancing. Subsequently, constructed evaluated have best accuracies 0.901 0.851, respectively. Finally, applied Daxaingling Tunnel, Cangling Zhongnanshan Tunnel shaft. results indicate that obtained levels consistent engineering records, reliable prediction.

Язык: Английский

Interpretable real-time monitoring of short-term rockbursts in underground spaces based on microseismic activities DOI Creative Commons
Mohammad Hossein Kadkhodaei, Ebrahim Ghasemi

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 6, 2025

In this study, two novel hybrid intelligent models were developed to evaluate the short-term rockburst using random forest (RF) method and meta-heuristic algorithms, whale optimization algorithm (WOA) coati (COA), for hyperparameter tuning. Real-time predictive of phenomenon created a database comprising 93 case histories, taking into account various microseismic parameters. The results indicated that WOA achieved highest overall performance in tuning RF model, outperforming COA. RF-WOA model accurately predicted occurrence with an accuracy 0.944. Additionally, precision, recall F1-score obtained as 0.950, 0.944 0.943, respectively, indicating proposed is robust predicting damage severity deep underground projects. Subsequently, Shapley additive explanations (SHAP) was employed interpret explain prediction process assess influence input features based on model. showed three parameters including cumulative seismic energy, events, apparent volume have greatest impact events. This study provides interpretable transparent resource events real time. It can facilitate estimating project costs, selecting suitable support system, identifying essential ways limit danger rockburst.

Язык: Английский

Процитировано

1
Comparative analysis and application of rockburst prediction model based on secretary bird optimization algorithm DOI Creative Commons
Ten-Fang Yang, Xinqiang Gao,

Lichuan Wang

и другие.

Frontiers in Earth Science, Год журнала: 2024, Номер 12

Опубликована: Дек. 16, 2024

The accurate rockburst prediction is crucial for ensuring the safety of underground engineering construction. Among various methods, machine learning-based can better solve nonlinear relationship between rockbursts and influencing factors thus has great potential applications. However, current research often faces certain challenges related to feature selection indices poor model optimization performance. This study compiled 342 cases from domestic international sources construct an initial database. In order determine relevant indicators, a method based on ReliefF-Kendall was proposed. database equalized visualized using Adasyn t-SNE algorithms. Five models [support vector (SVM), least-squares support (LSSVM), kernel extreme learning (KELM), Random Forest (RF), XGBoost] were established by employing Secretary Bird Optimization (SBO) algorithm 5-fold cross-validation optimize optimal selected comprehensive assessment generalization ability (accuracy, kappa, precision, recall, F1-score) stability (average accuracy). reliability proposed selection, optimization, data balancing methods verified comparing with other methods. results indicate that PSO-SVM demonstrated superior accuracy performance; reach 81.4% (optimal) 80.1% (average). main affecting occurrence are W et , maximum tangential stress ( MTS ), D uniaxial compressive strength UCS ). Finally, applied cases, achieving 90% verifying its applicability.

Язык: Английский

Процитировано

3
Event recognition technology and short-term rockburst early warning model based on microseismic monitoring and ensemble learning DOI Creative Commons
Zibin Li, Dengpan Qiao, Tianyu Yang

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 28, 2025

Язык: Английский

Процитировано

0
Performance Evaluation of Hybrid PSO-BPNN-AdaBoost and PSO-BPNN-XGBoost Models for Rockburst Prediction with Imbalanced Datasets DOI Creative Commons
Shujian Li, Pengpeng Lu, Weizhang Liang

и другие.

Applied Sciences, Год журнала: 2024, Номер 14(24), С. 11792 - 11792

Опубликована: Дек. 17, 2024

The rockburst hazard is a primary geological disaster endangering the environment in underground engineering. Due to complexity of mechanism, traditional methods are insufficient predict objectively, especially when dealing with an imbalanced dataset. To address this issue, hybrid models PSO-BPNN-AdaBoost and PSO-BPNN-XGBoost were developed hazards study. First, dataset 266 cases was constructed, containing six indicators: maximum tangential stress, uniaxial compressive strength, tensile elastic deformation energy index, stress brittleness coefficient strength. Then, original oversampled using synthetic minority oversampling technique (SMOTE) for balancing. Subsequently, constructed evaluated have best accuracies 0.901 0.851, respectively. Finally, applied Daxaingling Tunnel, Cangling Zhongnanshan Tunnel shaft. results indicate that obtained levels consistent engineering records, reliable prediction.

Язык: Английский

Процитировано

2