Abnormally High Water Temperature Prediction Using LSTM Deep Learning Model
-
Title
- Abnormally High Water Temperature Prediction Using LSTM Deep Learning Model
-
Author(s)
- Choi, Hey Min; Kim, Min Kyu; Yang, Hyun
- KIOST Author(s)
- Kim, Min Kyu(김민규)
-
Alternative Author(s)
- 최혜민; 김민규; 양현
-
Publication Year
- 2020-10-24
-
Abstract
- Recently, abnormally high water temperature (AHWT) phenomena are occurring more often due to the global warming and its impact. These phenomena have damaged extensively to the maritime economy around the southern coast of Korea by causing a mass stranding of farmed fish. Also, these phenomena cause an illness by exacerbating the propagation of Vibrio pathogens. To mitigate damages by AHWT phenomena, it is necessary to respond as quickly as possible or predict them in advance. In this study, therefore, we proposed a deep learning-based methodology to predict the occurrences of AHWT phenomena using the long short-term memory (LSTM) model. First, a LSTM model was trained using the satellite-derived water temperature data over the past fifteen years. Then, the water temperatures after three days were estimated using the trained LSTM model. In a performance evaluation, the model achieved results of 2.099 and 0.535 in terms of mean absolute percentage error (MAPE) and root mean square error (RMSE), respectively. Second, we developed a decision algorithm based on the Markov state transition in order to predict the AHWT occurrence probability. As a result, we obtained 0.714 of F1 score for predicting AHWT phenomena in case of the southern coast of Korea. We are expecting that this study contributes to protecting the aquaculture industry and mitigating the damages by AHWT phenomena.
-
URI
- https://sciwatch.kiost.ac.kr/handle/2020.kiost/38970
-
Bibliographic Citation
- The 9th International Multi-Conference on Engineering and Technology Innovation (IMETI2020), 2020
-
Publisher
- Taiwan Association of Engineering and Technology Innovation
-
Type
- Conference
-
Language
- English
- Files in This Item:
-
There are no files associated with this item.
Items in ScienceWatch@KIOST are protected by copyright, with all rights reserved, unless otherwise indicated.