Title: Determining the water level fluctuations of Lake Van through the integrated machine learning methods

Authors: Uğur Serencam; Ömer Ekmekcioğlu; Eyyup Ensar Başakın; Mehmet Özger

Addresses: Department of Civil Engineering, Marmara University, Istanbul, Turkey ' Hydraulics Division, Civil Engineering Department, Istanbul Technical University, Istanbul, Turkey ' Hydraulics Division, Civil Engineering Department, Istanbul Technical University, Istanbul, Turkey ' Hydraulics Division, Civil Engineering Department, Istanbul Technical University, Istanbul, Turkey

Abstract: Determining the lake levels is of paramount importance considering the environmental challenges encountered due to the global warming. The purpose of this study is to predict water level fluctuation of Lake Van using extreme gradient boosting (XGBoost). In addition, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method was adopted to the proposed model. The gravitational search algorithm (GSA) was utilised to tune the hyperparameters of XGBoost and the genetic algorithm (GA) and particle swarm optimisation (PSO) were used for benchmarking. The results showed that GSA-CEEMDAN-XGBoost model outperformed its counterparts, i.e., GA-CEEMDAN-XGBoost and PSO-CEEMDAN-XGBoost, according to the performance metrics.

Keywords: tree-based ensemble machine learning; water level forecast; signal processing; Lake Van; Mann-Whitney U test; hyperparameter optimisation; XGBoost.

DOI: 10.1504/IJGW.2022.123278

International Journal of Global Warming, 2022 Vol.27 No.2, pp.123 - 142

Received: 12 Aug 2021
Accepted: 22 Oct 2021
Published online: 07 Jun 2022 *

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