通过插值与稀疏化提升时序预测精度
Less is more: Embracing sparsity and interpolation with Esiformer for time series forecasting
- 先对原始数据插值降噪，再用稀疏前馈网络增强模型表达力
- 在多变量时序上比PatchTST降低6.5% MSE和5.8% MAE
- 适合噪声大、周期性弱的真实场景时序预测任务
时序预测在众多实际应用中至关重要。但真实世界生成的时序数据常具高方差与大量噪声，难以捕捉内在周期模式，显著影响预测精度。为此，我们提出Esiformer，通过在原始数据上进行插值，降低整体方差并缓解噪声影响。同时，我们在基础Transformer中引入稳健的稀疏前馈网络（Sparse FFN），有效增强模型表征能力，并保持良好鲁棒性，避免了标准实现中的过拟合风险。在多个具有挑战性的真实数据集上评估显示，该方法优于领先模型PatchTST，多变量时序预测中均方误差（MSE）降低6.5%，平均绝对误差（MAE）降低5.8%。代码已开源：https://github.com/yyg1282142265/Esiformer/tree/main。
原文摘要 · Abstract (English)
Time series forecasting has played a significant role in many practical fields. But time series data generated from real-world applications always exhibits high variance and lots of noise, which makes it difficult to capture the inherent periodic patterns of the data, hurting the prediction accuracy significantly. To address this issue, we propose the Esiformer, which apply interpolation on the original data, decreasing the overall variance of the data and alleviating the influence of noise. What's more, we enhanced the vanilla transformer with a robust Sparse FFN. It can enhance the representation ability of the model effectively, and maintain the excellent robustness, avoiding the risk of overfitting compared with the vanilla implementation. Through evaluations on challenging real-world datasets, our method outperforms leading model PatchTST, reducing MSE by 6.5% and MAE by 5.8% in multivariate time series forecasting. Code is available at: https://github.com/yyg1282142265/Esiformer/tree/main.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。