Answers are grounded in this paper and cite their sources.
This paper introduces IGDS-ASDE, a two-stage ensemble approach for time series forecasting that addresses the challenges of ensemble member selection and adaptive combination. This approach comprises two components: Importance Greedy Diversity Selection (IGDS) and Attention Score-based Dynamic Ensemble (ASDE). IGDS is used to eliminate redundant and underperforming ensemble members from the candidate pool and enhance diversity. Then ASDE uses attention mechanism in a meta-learning framework to adaptively compute ensemble member weights for the final prediction based on historical and predicted future performance. Experimental results demonstrate that the different versions of IGDS-ASDE performed well and achieved significant improvements over advanced ensemble and individual baselines used for comparison, across seven time series datasets.
No takes yet. Share an insight, caveat, or question.
Yao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: