Evolutionary optimization of neural network for landslide susceptibility assessment

TAŞKANAT T., Karkınlı A. E., Kesikoğlu A., BEŞDOK E.

Stochastic Environmental Research and Risk Assessment, 2025 (SCI-Expanded, Scopus) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1007/s00477-025-03118-6
  • Dergi Adı: Stochastic Environmental Research and Risk Assessment
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Agricultural & Environmental Science Database, CAB Abstracts, Compendex, Environment Index, Geobase, Index Islamicus, Pollution Abstracts, zbMATH, Civil Engineering Abstracts
  • Anahtar Kelimeler: Evolutionary algorithm, Geographic information systems, Landslide susceptibility assessment, Neural network, Optimization
  • Kayseri Üniversitesi Adresli: Evet

Özet

This research investigates the optimization of artificial neural networks (ANNs) for landslide susceptibility assessment in the Özvatan district of Kayseri province, Turkey. It expands upon existing approaches by incorporating the novel evolutionary algorithms : Backtracking Search Algorithm (BSA), Bernstein Search Differential Evolution Algorithm (BSD), and Weighted Differential Evolution Algorithm (WDE), which require minimal user-defined parameters and offer robust convergence, alongside established methods, such as the Particle Swarm Optimization Algorithm (PSO) and Differential Evolution Algorithm (DE). The susceptibility assessment utilizes a spatial database comprising 100 historical landslide polygons and thirteen landslide conditioning factors. These factors encompass elevation, slope characteristics (slope, aspect, and plan and profile curvatures), topographic indexes (topographic wetness index (TWI), stream power index (SPI), sediment transport index (STI), and topographic position index (TPI)), distance to drainage network, normalized difference vegetation index (NDVI), and distance to lineaments. The performance of the proposed models (ANN, BSA-ANN, BSD-ANN, DE-ANN, PSO-ANN, and WDE-ANN) is evaluated using a comprehensive suite of metrics, including mean squared error (MSE), area under the receiver operating characteristic curve (AUROC), accuracy, precision, recall, and F1 score. The findings show that the BSD-ANN model achieved the highest performance, with an AUROC of 0.771 and an F1 Score of 0.782 in the testing phase, significantly outperforming the standalone ANN model.