Research map / Time series / Causal and generative time series
MixSeq: Connecting Macroscopic Time Series Forecasting with Microscopic Time Series Data
Zhibo Zhu, Ziqi Liu, Ge Jin, Zhiqiang Zhang, Lei Chen, JUN ZHOU, Jianyong Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Cluster microscopic series with MixSeq, then forecast and aggregate them instead of forecasting the macro series directly | ConvTrans on Rossmann · Rossmann | SMAPE: 0.1028(0.0091) -> 0.0961(0.0019) | Improved |
| Cluster microscopic series with MixSeq, then forecast and aggregate them | ConvTrans on M5 · M5 | SMAPE: 0.0245(0.0024) -> 0.0227(0.0006) | Improved |
| Cluster microscopic series with MixSeq, then forecast and aggregate them | ConvTrans on Wiki · Wiki | SMAPE: 0.0466(0.0001) -> 0.0440(0.0010) | Improved |
| Use MixSeq clustering rather than macro forecasting | ConvTrans on Rossmann · Rossmann | R0.5 loss: 0.1861 -> 0.1847 | Improved |
| Use MixSeq clustering rather than macro forecasting | ConvTrans on Rossmann · Rossmann | R0.9 loss: 0.0822 -> 0.0813 | Improved |
| Use MixSeq clustering rather than the strongest listed alternative | ConvTrans on M5 · M5 | R0.5 loss: 0.0497 -> 0.0460 | Improved |
| Use MixSeq clustering rather than the strongest listed alternative | ConvTrans on M5 · M5 | R0.9 loss: 0.0257 -> 0.0238 | Improved |
| Use MixSeq clustering rather than the strongest listed alternative | ConvTrans on Wiki · Wiki | R0.5 loss: 0.0961 -> 0.0901 | Improved |
| Use MixSeq clustering rather than the strongest listed alternative | ConvTrans on Wiki · Wiki | R0.9 loss: 0.0531 -> 0.0516 | Improved |
| Use a flexible Seq2seq mixture model rather than an ARMA mixture | Synthetic DeepAR time-series clustering · DeepAR synthetic data, 2 clusters | Rand Index: 0.7995(0.2734) -> 0.9986(0.0003) | Improved |
| Aggregate time series by their known generating component before forecasting | Forecasting macroscopic series in toy mixtures · GP and ARMA mixtures, 3 and 5 clusters | SMAPE: 0.0263 -> 0.0210; 0.0242 -> 0.0198; 0.5870 -> 0.3590; 0.5940 -> 0.3840 | Improved |
| Use 3 clusters, matching the ground-truth cluster count, rather than 2 (ablation) | MixSeq with ARMA forecasting · Synthetic ARMA forecasting | SMAPE: 0.774 -> 0.731 | Improved |
Closest papers by meaning
- xLSTM-Mixer: Multivariate Time Series Forecasting by Mixing via Scalar Memories (NeurIPS 2025 · similarity 0.86)
- TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024 · similarity 0.85)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.85)
- MSGNet: Learning Multi-Scale Inter-series Correlations for Multivariate Time Series Forecasting (AAAI 2024 · similarity 0.84)
- Addressing Spatial-Temporal Heterogeneity: General Mixed Time Series Analysis via Latent Continuity Recovery and Alignment (NeurIPS 2024 · similarity 0.84)
- Chimera: Effectively Modeling Multivariate Time Series with 2-Dimensional State Space Models (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
