Time Evidence Fusion Network: Multi-source View in Long-Term Time Series Forecasting

10 May 2024  ·  Tianxiang Zhan, Yuanpeng He, Zhen Li, Yong Deng ·

In real-world scenarios, time series forecasting often demands timeliness, making research on model backbones a perennially hot topic. To meet these performance demands, we propose a novel backbone from the perspective of information fusion. Introducing the Basic Probability Assignment (BPA) Module and the Time Evidence Fusion Network (TEFN), based on evidence theory, allows us to achieve superior performance. On the other hand, the perspective of multi-source information fusion effectively improves the accuracy of forecasting. Due to the fact that BPA is generated by fuzzy theory, TEFN also has considerable interpretability. In real data experiments, the TEFN partially achieved state-of-the-art, with low errors comparable to PatchTST, and operating efficiency surpass performance models such as Dlinear. Meanwhile, TEFN has high robustness and small error fluctuations in the random hyperparameter selection. TEFN is not a model that achieves the ultimate in single aspect, but a model that balances performance, accuracy, stability, and interpretability.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Time Series Forecasting Electricity (192) TEFN MSE 0.197 # 4
MAE 0.276 # 1
Time Series Forecasting Electricity (336) TEFN MSE 0.212 # 4
MAE 0.292 # 1
Time Series Forecasting Electricity (720) TEFN MSE 0.253 # 5
MAE 0.325 # 1
Time Series Forecasting Electricity (96) TEFN MSE 0.197 # 5
MAE 0.273 # 1
Time Series Forecasting ETTh1 (192) Multivariate TEFN MSE 0.433 # 11
MAE 0.419 # 4
Time Series Forecasting ETTh1 (336) Multivariate TEFN MSE 0.475 # 11
MAE 0.441 # 7
Time Series Forecasting ETTh1 (720) Multivariate TEFN MSE 0.475 # 10
MAE 0.464 # 10
Time Series Forecasting ETTh1 (96) Multivariate TEFN MSE 0.383 # 8
MAE 0.391 # 4
Time Series Forecasting ETTh2 (192) Multivariate TEFN MSE 0.375 # 10
MAE 0.392 # 3
Time Series Forecasting ETTh2 (336) Multivariate TEFN MSE 0.423 # 11
MAE 0.434 # 5
Time Series Forecasting ETTh2 (720) Multivariate TEFN MSE 0.434 # 8
MAE 0.446 # 8
Time Series Forecasting ETTh2 (96) Multivariate TEFN MSE 0.288 # 10
MAE 0.337 # 4
Time Series Forecasting ETTm1 (192) Multivariate TEFN MSE 0.381 # 4
MAE 0.383 # 2
Time Series Forecasting ETTm1 (336) Multivariate TEFN MSE 0.414 # 3
MAE 0.404 # 2
Time Series Forecasting ETTm1 (720) Multivariate TEFN MSE 0.475 # 3
MAE 0.438 # 2
Time Series Forecasting ETTm1 (96) Multivariate TEFN MSE 0.343 # 4
MAE 0.367 # 2
Time Series Forecasting ETTm2 (192) Multivariate TEFN MSE 0.381 # 4
MAE 0.304 # 2
Time Series Forecasting ETTm2 (336) Multivariate TEFN MSE 0.307 # 3
MAE 0.343 # 2
Time Series Forecasting ETTm2 (720) Multivariate TEFN MSE 0.407 # 3
MAE 0.398 # 2
Time Series Forecasting ETTm2 (96) Multivariate TEFN MSE 0.181 # 4
MAE 0.264 # 2
Time Series Forecasting Weather (192) TEFN MSE 0.227 # 8
MAE 0.262 # 1
Time Series Forecasting Weather (336) TEFN MSE 0.279 # 6
MAE 0.298 # 1
Time Series Forecasting Weather (720) TEFN MSE 0.352 # 6
MAE 0.344 # 1
Time Series Forecasting Weather (96) TEFN MSE 0.182 # 7
MAE 0.227 # 1

Methods


No methods listed for this paper. Add relevant methods here