Table 1 Baseline models used for comparison.

From: Ocean environment prediction methods based on deep learning and spatiotemporal feature fusion

Method

Abbreviation

Relation to MSSTN

Neural Network + Autoregressive Integrated Moving Average (ARIMA)26

Mod1

Linear-nonlinear fusion

Long Short-Term Memory Neural Network (LSTM) 27

Mod2

Temporal module

Convolutional Gated Recurrent Unit (CNN-GRU)28

Mod3

Spatial conv vs. MSGCN

Time-series Graph Network (TSGN)29

Mod4

Static vs. dynamic graphs

GCN-LSTM30

Mod5

Single-graph vs. multi-graph

Spatiotemporal Graph Convolutional Network (ST-GCN)31

Mod6

Spatial granularity

Spatiotemporal coupled attention network (STCANET) 32

Mod7

Unidirectional vs. BDTB

The model proposed in this paper

MSSTN

 
  1. This table summarizes the models selected as baselines to evaluate the performance of the proposed MSSTN. The baselines include statistical approaches (e.g., ARIMA), temporal deep learning models (e.g., LSTM), and spatiotemporal deep learning architectures (e.g., GCN-LSTM, STCANET). By covering these three categories, the comparison ensures a comprehensive evaluation against both traditional and state-of-the-art methods.
Back to article page