Global Temperature Prediction Models Based on ARIMA and LSTM

Yu, Yue; Xie, Yi; Tao, Zui; Ju, Hongmei; Wang, Meiling

doi:10.1007/978-981-99-7549-5_22

Cited by 1 publication

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Xi' = Ms (Xi) ⊗ Xi Yi = Ttrans (Ms (Xi') ⊗ Xi') Aj' = Mt (Aj) ⊗ Aj Bj = Ttrans (Mt (Aj') Aj') Lk = Yi + Bj (11) The symbol ⊗ represents element-wise multiplication. The final refined output, denoted as Lk, is then passed into the next module in the model.…”

Section: Model Optimizationmentioning

confidence: 99%

“…It was first proposed by Hochreiter (1997) and refined and popularized by many people later. With the rapid development of deep learning, Long Short-Term Memory (LSTM) [8][9][10][11] is also widely used in trajectory sequence prediction tasks. The prediction methods based on deep learning avoid unknown degradation, extract effective information from monitoring data, depict the nonlinear relationship between feature information and life, and can track and predict the trajectory more accurately [12][13][14].…”

Section: Introductionmentioning

confidence: 99%

“…In summary, due to the complexity of trajectory prediction, the existing intelligent predictions of the ocean mainly focus on aspects such as seawater environment and marine disasters [8,9,11,12,14,[16][17][18][19][20][21][22][23][24][25]. Currently, Bayesian model-based methods are not applied in trajectory research [26].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Argo Buoy Trajectory Prediction: Multi-Scale Ocean Driving Factors and Time–Space Attention Mechanism

Ning,

Zhang,

Zhang

et al. 2024

JMSE

View full text Add to dashboard Cite

The Array for Real-time Geostrophic Oceanography (Argo) program provides valuable data for maritime research and rescue operations. This paper is based on Argo historical and satellite observations, and inverted sea surface and submarine drift trajectories. A neural network method was developed to predict the position of Argo buoys, improving target tracking and emergency support capabilities. Based on a deep learning framework using a Simple Recurrent Unit (SRU), a new Time–Space Feature Fusion Method based on an Attention Mechanism (TSFFAM) model was constructed. The TSFFAM mechanism can predict the target trajectory more accurately, avoiding the disadvantages of traditional Long Short-Term Memory (LSTM) models, which are time consuming and difficult to train. The TSFFAM model is able to better capture multi-scale ocean factors, leading to more accurate and efficient buoy trajectory predictions. In addition, it aims to shed light on the mechanism of the joint multi-element and multi-scale effects of laminar and surface currents on multi-scale ocean factors, thereby deepening our understanding of the multi-element and multi-scale interactions in different spatio-temporal regimes of the ocean. Experimental verification was conducted in the Pacific Ocean using buoy trajectory data, and the experimental results showed that the buoy trajectory prediction models proposed in this paper can achieve high prediction accuracy, with the TSFFAM model improving the accuracy rate by approximately 20%. This research holds significant practical value for the field of maritime studies, precise rescue operations, and efficient target tracking.

show abstract

Section: Model Optimizationmentioning

confidence: 99%