We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
Multivariate Prediction Framework of Ocean Sensing Data Assisting Red Tide Early warning.
- Authors
Sun, Xiaochuan; Cao, Difei; Fan, Xianchuang; Li, Zhigang; Li, Yingqi
- Abstract
In red tide early waring, multivariate ocean sensing data prediction has been the current technical mainstream. This prediction pattern can effectively capture the impact information of nonpredictive ocean environmental factors on the predictive target one, thus achieving a good nonlinear approximation. However, the temporal-spatial change of these impact information is completely ignored in practical applications. That is to say, these multivariate factors are generally determined by experience, which is typically not conducive to accurate prediction of targeted ocean series. To address this issue, this paper develops a novel multivariate ocean sensing data prediction framework combining cross recurrence theory (CRT) and long short-term memory neural network (LSTM), called CRT-mLSTM. In structure, this framework is composed of three functional parts of visualization, qualitative correlation analysis and multivariate prediction. CRT is employed to analyze the correlations between targeted chlorophyll a (Chla) series and other ocean environmental factors, which determines the optimal multivariate inputs for the LSTM-based predictor. On the collected real-world ocean sensing data of different sea areas, experimental results show that the proposed CRT-mLSTM significantly outperforms the state-of-the-art approaches in prediction performance. Remarkably, our proposal can obtain the RMSE, R 2 and MAE of 5.08% , 98.82% and 3.88% on the Chla dataset of see area B, respectively. Such superiority is further verified by the visualized consistency of statistical distribution.
- Subjects
RED tide; DISTRIBUTION (Probability theory); STATISTICAL correlation; MULTIVARIATE analysis; DATA visualization; OCEAN
- Publication
Arabian Journal for Science & Engineering (Springer Science & Business Media B.V. ), 2023, Vol 48, Issue 8, p10963
- ISSN
2193-567X
- Publication type
Article
- DOI
10.1007/s13369-023-07788-8