A novel method for landslide displacement prediction by integrating advanced computational intelligence algorithms.

Area of Science:

• Geotechnical Engineering
• Computational Intelligence
• Earthquake Engineering

Background:

• Landslide displacement prediction is vital for early warning systems.
• Conventional methods require extensive data, limiting their applicability.
• Accurate prediction with limited data remains a challenge.

Purpose of the Study:

• To propose and validate a novel hybrid computational intelligence method for landslide displacement prediction.
• To enhance prediction accuracy and stability using limited monitoring data.
• To compare the proposed method against existing techniques.

Main Methods:

• Integrating Wavelet Transform (WT) for signal decomposition, Artificial Bees Colony (ABC) for parameter optimization, and Kernel-based Extreme Learning Machine (KELM) for prediction.
• Decomposing total landslide displacement into multiple frequency sub-sequences using WT.
• Optimizing KELM parameters with ABC and predicting each sub-sequence separately before summation.

Main Results:

• The proposed WT-ABC-KELM method significantly improved prediction accuracy compared to WT-ELM, ABC-KELM, ELM, and SVM.
• Decomposing displacement into sub-sequences and predicting them individually enhanced overall accuracy.
• The ABC-KELM algorithm demonstrated superior prediction capacity, followed by ELM and SVM.

Conclusions:

• The integrated WT-ABC-KELM approach offers an effective solution for accurate landslide displacement prediction with limited data.
• The method exhibits excellent accuracy and stability, making it suitable for early warning systems.
• Decomposition and separate prediction of displacement sub-sequences is a beneficial strategy for improving forecast models.