Napredna pretraga

Pregled bibliografske jedinice broj: 1027668

LAKE LEVEL PREDICTION USING LONG SHORT-TERM MEMORY RECURRENT NEURAL NETWORKS


Hrnjica, Bahrudin; Bonacci, Ognjen
LAKE LEVEL PREDICTION USING LONG SHORT-TERM MEMORY RECURRENT NEURAL NETWORKS // 12th INTERNATIONAL SCIENTIFIC CONFERENCE "DEVELOPMENT AND MODERNIZATION OF PRODUCTION"
Bihač: University of Bihač, 2019. str. 274-279 (plenarno, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)


Naslov
LAKE LEVEL PREDICTION USING LONG SHORT-TERM MEMORY RECURRENT NEURAL NETWORKS

Autori
Hrnjica, Bahrudin ; Bonacci, Ognjen

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
12th INTERNATIONAL SCIENTIFIC CONFERENCE "DEVELOPMENT AND MODERNIZATION OF PRODUCTION" / - Bihač : University of Bihač, 2019, 274-279

Skup
12th INTERNATIONAL SCIENTIFIC CONFERENCE "DEVELOPMENT AND MODERNIZATION OF PRODUCTION"

Mjesto i datum
Sarajevo, Bosna i Hercegovina, 18-20.09.2019

Vrsta sudjelovanja
Plenarno

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Time series, lake level, LSTM, RNN, karst hydrology

Sažetak
N this paper, the artificial neural network was used to develop a month ahead prediction model for Vrana lake level. Vrana lake is located on the island of Cres in the Croatian part of the Adriatic Sea. It is one of the largest natural freshwater sources on Mediterranean islands. In order to develop a reliable and accurate prediction model the Long Short-Term Memory (LSTM) recurrent neural network was used. The model was trained on time series data which represent an average monthly level measured in the last 40 years. The data were split on training, validation, and testing set in order to provide a reliable foundation for the model training, evaluation and model prediction. Once the model is trained, the evaluation and testing were performed in order to prove the model's accuracy and generalizability. The results showed that using the LSTM recurrent neural network, can be obtain models better that models calculated using simple feed-forward neural network. The results were shown the lake is facing a dangerous decreasing level caused by several factors described in the paper.

Izvorni jezik
Engleski

Znanstvena područja
Građevinarstvo



POVEZANOST RADA


Projekt / tema
083-0831510-1511 - Proučavanje ekstremnih hidroloških situacija i vodnih rizika u kršu (Ognjen Bonacci, )

Ustanove
Fakultet građevinarstva, arhitekture i geodezije, Split

Autor s matičnim brojem:
Ognjen Bonacci, (4434)