|
This article is cited in 2 scientific papers (total in 2 papers)
MULTISCALE MODELING FOR INFORMATION CONTROL AND PROCESSING
Finding the optimal machine learning model for flood prediction on the Amur river
N. E. Aleksandrova, D. N. Ermakovab, N. M. Aziza, O. Yu. Kazenkovabc a Engineering Academy of the Peoples’ Friendship University (RUDN University)
b Polyus Research and Development Institute named after M. F. Stel'makh, Moscow
c Moscow State University of Technologies and Managements
Abstract:
Water-related natural disasters are among the most devastating and are responsible for 72% of the total economic damage caused by natural disasters, and due to climate change, their number will only increase. In Russia, river floods are the main such disaster. The purpose of this research work is to determine the best machine learning method for predicting floods on the Amur River, where they cause significant damage to the population and economy of the region. The study was undertaken with the aim of improving flood forecasting methods for the subsequent use of the study results in solving management problems in response to floods. The study considers the practical aspects of implementing a forecasting system, so the 3 most popular machine learning methods were studied: linear regression, neural network and gradient boosting, because these methods have a developed ecosystem of auxiliary solutions and are widely known in the professional community. The research methodology was aimed at achieving maximum comparability of results. Among the algorithms tested, gradient boosting over trees in the implementation of Catboost demonstrated the best quality. The results of the study are also applicable to other rivers, for which the amount of data is comparable to that of the Amur.
Keywords:
disaster management, floods forecasting, Amur River, machine learning.
Received: 26.04.2022
Citation:
N. E. Aleksandrov, D. N. Ermakov, N. M. Aziz, O. Yu. Kazenkov, “Finding the optimal machine learning model for flood prediction on the Amur river”, Comp. nanotechnol., 9:2 (2022), 11–20
Linking options:
https://www.mathnet.ru/eng/cn371 https://www.mathnet.ru/eng/cn/v9/i2/p11
|
Statistics & downloads: |
Abstract page: | 84 | Full-text PDF : | 60 | References: | 1 |
|