Research map / Time series / Long-term forecasting
ProtoTS: Learning Hierarchical Prototypes for Explainable Time Series Forecasting
Ziheng Peng, Shijie Ren, Xinyue Gu, Linxiao Yang, Xiting Wang, Liang Sun
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 |
|---|---|---|---|
| add bottleneck layer (ablation) | ProtoTS · LOF (avg of 4 regions) | MAE: 0.143 -> 0.106 | Improved |
| add multi-channel embedding (vs single channel) (ablation) | ProtoTS · LOF (avg of 4 regions) | MAE: 0.117 -> 0.106 | Improved |
| add hierarchical prototype structure (vs flat) (ablation) | ProtoTS · LOF (avg of 4 regions) | MAE: 0.110 -> 0.106 | Improved |
| add bottleneck layer (ablation) | ProtoTS · LOF (avg of 4 regions) | MSE: 0.049 -> 0.031 | Improved |
| increase number of root prototypes from 6 to 18 (ablation) | ProtoTS · LOF | MAE | Improved |
| increase root prototypes beyond a threshold (ablation) | ProtoTS · LOF | MAE | No clear effect |
| reduce training data from 100% to 50% (ablation) | ProtoTS vs TimeXer/iTransformer/TiDE · LOF PC | MSE | No clear effect |
| expert splits Spring Festival prototype | ProtoTS · LOF RE, Spring Festival | MSE: 0.099 -> 0.090 | Improved |
| use prototype-based hierarchical model vs best baseline | ProtoTS · LOF | MSE: 48.3% reduction | Improved |
| use ProtoTS instead of TFT/NBEATSx for explanations | ProtoTS · User study | SUS score: 73.36 vs 29.74 (TFT), 38.66 (NBEATSx) | Improved |
| ProtoTS compute cost | ProtoTS · LOF | training ms/iter: 40.8 vs 19.5 (TimeXer), 12.5 (iTrans), 11.1 (TiDE) | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 4 of 4 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SynTSBench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series (NeurIPS 2025 · similarity 0.89)
- Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization (ICML 2025 · similarity 0.87)
- HDT: Hierarchical Discrete Transformer for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.87)
- TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024 · similarity 0.87)
- TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop (NeurIPS 2025 · similarity 0.86)
- Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024 · similarity 0.86)
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.
