Article: Hyperparameters optimisation of ensemble classifiers and its application for landslide hazards classification Journal: International Journal of Modelling, Identification and Control (IJMIC) 2022 Vol.40 No.2 pp.158 - 175 Abstract: Along with assessing the hazards of landslides taking into consideration the faced difficulties and the consumed time when determining the algorithms configurations and parameters manually, the primary aspiration of this study is to optimise the parameters of two ensemble-based machine learning algorithms using particle swarm optimisation, genetic algorithm, and Bayesian optimisation so that the optimised algorithms can identify and classify landslides more efficiently and accurately. Random forest classifier and XGBoost models were used and the ADASYN was implemented to overcome the shortage of imbalanced data. In the experiments, it was clearly shown that the hypered ensemble-based models along with the PSO and GA successfully surpassed the single models on classifying the landslides' triggers, sizes, and types. The experimental results demonstrated that the hyperparameters optimisation can greatly improve the accuracy of the ensemble classifiers, thus it can provide accurate classification results and decision support for the disaster prevention and mitigation management departments. Inderscience Publishers - linking academia, business and industry through research

Title: Hyperparameters optimisation of ensemble classifiers and its application for landslide hazards classification

Authors: Jiuyuan Huo; Hamzah Murad Mohammed Al-Neshmi

Addresses: School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China; National Cryosphere Desert Data Center (NCDC), Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China ' School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China

Abstract: Along with assessing the hazards of landslides taking into consideration the faced difficulties and the consumed time when determining the algorithms configurations and parameters manually, the primary aspiration of this study is to optimise the parameters of two ensemble-based machine learning algorithms using particle swarm optimisation, genetic algorithm, and Bayesian optimisation so that the optimised algorithms can identify and classify landslides more efficiently and accurately. Random forest classifier and XGBoost models were used and the ADASYN was implemented to overcome the shortage of imbalanced data. In the experiments, it was clearly shown that the hypered ensemble-based models along with the PSO and GA successfully surpassed the single models on classifying the landslides' triggers, sizes, and types. The experimental results demonstrated that the hyperparameters optimisation can greatly improve the accuracy of the ensemble classifiers, thus it can provide accurate classification results and decision support for the disaster prevention and mitigation management departments.

Keywords: landslide; optimisation; random forest; extreme gradient boosting; XGBoost; adaptive synthetic; ADASYN.

DOI: 10.1504/IJMIC.2022.124724

International Journal of Modelling, Identification and Control, 2022 Vol.40 No.2, pp.158 - 175

Received: 30 Jul 2021
Accepted: 24 Sep 2021
Published online: 08 Aug 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Hyperparameters optimisation of ensemble classifiers and its application for landslide hazards classification

Keep up-to-date