Forecasting results obtained on real-world datasets using the OCNS. (IMAGE)
Caption
(a) Comparison of the forecasting performance and parameter efficiency of OCNS, OCNS+, and ten baseline methods across multiple real-world datasets. The mean squared errors (MSEs) are presented alongside the number of parameters required by each method. Due to exceeding the 24-hour training limit, MSEs for the Time-LLMb method on WTH2, ECL, and SMD datasets are unavailable. Panels (b)-(g) visually compare the forecasting outcomes of OCNS with those of the other methods on the ETTh1, ETTh2, WTH1, WTH2, ECL, and ETTm1 datasets, respectively.
Credit
©Science China Press
Usage Restrictions
Use with credit.
License
Original content