@article {
author = {Ghaffari-Razin, Seyyed Reza and Hooshangi, Navid and Heydari, Fateme and Safari, Hanie},
title = {Efficiency of adaptive neuro-fuzzy inference system in estimation of Earth's velocity field},
journal = {Earth and Statistics},
volume = {2},
number = {1},
pages = {-},
year = {2023},
publisher = {Arak University of Technology},
issn = {2980-9339},
eissn = {2980-9339},
doi = {10.22034/jes.2023.556488.1006},
abstract = {The purpose of this paper is estimation of earth velocity field using adaptive neuro-fuzzy inference system (ANFIS) in north-west of Iran. For this purpose, observations of 22 GPS stations are selected from the Azerbaijan local network. At the first step, the velocity field of these stations is estimated with Bernese GNSS software. Then, observations of 20 stations used for training of ANFIS with back-propagation (BP) algorithm. It should be note that the input vector of ANFIS is considered latitude and longitude of the GPS stations and the output is the velocity field (Ve , Vn). After the training step, estimated velocity field with ANFIS in two test stations, are compared and evaluated with the velocity field obtained from GPS. For a more accurate evaluation of the new model, all results are compared with velocity field obtained from the Kriging model. To analyze the error of the models, relative error and root mean square error (RMSE) is used. The averaged relative error of ANFIS and Kriging models obtained in the two test stations for the eastern component of the velocity field is 8.61% and 18.99%, respectively. Also for northern component, the averaged relative error is 21.84% and 28.51%, respectively. The results show the high accuracy of ANFIS model compared to Kriging in estimation of velocity field.},
keywords = {},
title_fa = {Efficiency of adaptive neuro-fuzzy inference system in estimation of Earth's velocity field},
abstract_fa = {The purpose of this paper is estimation of earth velocity field using adaptive neuro-fuzzy inference system (ANFIS) in north-west of Iran. For this purpose, observations of 22 GPS stations are selected from the Azerbaijan local network. At the first step, the velocity field of these stations is estimated with Bernese GNSS software. Then, observations of 20 stations used for training of ANFIS with back-propagation (BP) algorithm. It should be note that the input vector of ANFIS is considered latitude and longitude of the GPS stations and the output is the velocity field (Ve , Vn). After the training step, estimated velocity field with ANFIS in two test stations, are compared and evaluated with the velocity field obtained from GPS. For a more accurate evaluation of the new model, all results are compared with velocity field obtained from the Kriging model. To analyze the error of the models, relative error and root mean square error (RMSE) is used. The averaged relative error of ANFIS and Kriging models obtained in the two test stations for the eastern component of the velocity field is 8.61% and 18.99%, respectively. Also for northern component, the averaged relative error is 21.84% and 28.51%, respectively. The results show the high accuracy of ANFIS model compared to Kriging in estimation of velocity field.},
keywords_fa = {},
url = {https://jes.arakut.ac.ir/article_703066.html},
eprint = {https://jes.arakut.ac.ir/article_703066_c55236ddc44d7ed802b0bfde3d56e91e.pdf}
}