Dynamic prediction of landslide life expectancy using ensemble system incorporating classical prediction models and machine learning

• DOI: 10.1016/j.gsf.2023.101758
• Publication Date: 2023-11-22T10:12:59Z
• Authors: Lei-Lei Liu; Hao-Dong Yin; Ting Xiao; Lei Huang; Yung-Ming Cheng
• Keywords: Ensemble forecasting; Predictive modelling; Ensemble Learning

With the development of landslide monitoring system, many attempts have been made to predict failure-time utilizing data displacements. Classical models (e.g., Verhulst, GM (1,1), and Saito models) that consider characteristics displacement determine investigated extensively. In practice, is continuously implemented with data-set updated, meaning predicted life expectancy (i.e., lag between time node at each instant conducting prediction) should be re-evaluated time. This manner termed "dynamic prediction". However, performances classical not discussed in context dynamic prediction yet. this study, such are firstly, disadvantages then reported, incorporating from four real landslides. Subsequently, a more qualified ensemble model proposed, where individual integrated by machine learning (ML)-based meta model. To evaluate quality under prediction, novel indicator 'discredit index (β)' higher value β indicates lower quality. It found Verhulst would produce results significantly β, while (1,1) indicate highest mean absolute error (MAE). Meanwhile, accurate than models. Here, performance decision tree regression (DTR)-based best among various ML-based