講演抄録/キーワード |
講演名 |
2018-08-27 10:00
Climate Forecasting by ConvLSTM on Segmented Region ○Ekasit Phermphoonphiphat(Osaka Univ.)・Tomohiko Tomita(Kumamoto Univ.)・Masayuki Numao・Ken-ichi Fukui(Osaka Univ.) AI2018-13 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
Recent years, climate forecasting techniques with machine learning have been developing to get high accuracy result. However, the forecasting area might contains local trends or specific characteristics of climate variables; therefore, training with only one model might not capture all of the local trends on entire area. Most researches on climate forecasting are focusing on just the small area such as country. Also spatial correlation needs to be considered to build forecasting model, the model that be able to maintain spatial correlation might have the better forecasting result. In this paper, we discussed availability of using segmented region to improve forecasting result and we compare non spatial correlation setup and spatial correlation model Convolutional Long-Short Term Memory (ConvLSTM). ConvLSTM be able to maintain spatial correlation with convolutional layer and maintain prior knowledge with forget-gate from LSTM. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Climate forecasting / Spatial correlation / ConvLSTM / / / / / |
文献情報 |
信学技報, vol. 118, no. 197, AI2018-13, pp. 1-6, 2018年8月. |
資料番号 |
AI2018-13 |
発行日 |
2018-08-20 (AI) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
AI2018-13 |