Table 1 Comparison of fusion model research in marine and environmental prediction.
From: Marine fishery resource dynamic prediction based on CNN-XGBoost fusion model
Study | Model architecture | Application domain | Dataset size | Key performance | Limitations addressed |
|---|---|---|---|---|---|
Li et al. (2017)27 | CNN + XGBoost | Social media popularity | 432 K images | Spearman’s ρ: 0.7406 | Feature extraction + classification |
Shi et al. (2024)28 | CNN-LSTM + Attention | Fishing effort prediction | 1,899 vessels | Error ratio: 5.6% | Spatiotemporal dynamics |
Xu et al. (2023)29 | LSTM | Tuna CPUE prediction | 15-year record | High accuracy | Temporal dependencies |
Zhang et al. (2017)21 | ConvLSTM | Sea surface temperature | Multi-year | - | Spatial-temporal patterns |
Thongsuwan et al. (2021)30 | CNN + XGBoost | Nuclear engineering | - | Superior to standalone | Classification tasks |
Titouni et al. (2025)31 | CNN + XGBoost | Wireless communication | Modulation dataset | Accuracy: 98.3% | Signal classification |
Our study | CNN + XGBoost Fusion | Marine fishery resources | 3.8 M records | RMSE: 2.847, R²: 0.846 | Temporal + heterogeneous features |
