arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2404.06198cs.LG
收藏

数据集相似性与多样性对时间序列预测中迁移学习成功的影响

The impact of data set similarity and diversity on transfer learning success in time series forecasting

Fraunhofer Institute for Integrated Circuits IIS(弗劳恩霍夫集成电路研究所);Fraunhofer Center for Applied Research on Supply Chain Services SCS(弗劳恩霍夫供应链服务应用研究中心);Christian-Albrechts-Universität zu Kiel(基尔基督教阿尔布雷希特大学);Otto-Friedrich-Universität Bamberg(班贝格奥托-弗里德里希大学);Vives University of Applied Science(维维斯应用科技大学);University of Haute-Alsace(上阿尔萨斯大学);Monash University(莫纳什大学)

Claudia Ehrig, Benedikt Sonnleitner, Ursula Neumann, Catherine Cleophas, Germain Forestier

AI总结:

本文系统研究了源-目标相似性与源多样性对时间序列迁移学习的影响,发现相似性降低偏差,多样性提升准确性与不确定性估计但增加偏差。

摘要:

预训练模型通过利用迁移学习,已成为提升目标数据集上时间序列预测效率和准确性的关键手段。尽管基准测试验证了模型在各类目标数据集上的泛化性能,但目前仍缺乏结构化的研究来提供相似性和多样性度量,以解释源数据与目标数据的哪些特征能够促成迁移学习的成功。本研究率先系统评估了源-目标相似性与源多样性对零样本和微调预测结果在准确性、偏差和不确定性估计方面的影响。我们使用在五个公开源数据集上预训练的神经网络,对五个目标数据集(包括真实世界批发数据)进行预测,从而研究这些动态关系。我们确定了两种基于特征的相似性和多样性度量,发现源-目标相似性能够降低预测偏差,而源多样性则能提升预测准确性和不确定性估计,但会增加偏差。

英文摘要:

Pre-trained models have become pivotal in enhancing the efficiency and accuracy of time series forecasting on target data sets by leveraging transfer learning. While benchmarks validate the performance of model generalization on various target data sets, there is no structured research providing similarity and diversity measures to explain which characteristics of source and target data lead to transfer learning success. Our study pioneers in systematically evaluating the impact of source-target similarity and source diversity on zero-shot and fine-tuned forecasting outcomes in terms of accuracy, bias, and uncertainty estimation. We investigate these dynamics using pre-trained neural networks across five public source datasets, applied to forecasting five target data sets, including real-world wholesales data. We identify two feature-based similarity and diversity measures, finding that source-target similarity reduces forecasting bias, while source diversity improves forecasting accuracy and uncertainty estimation, but increases the bias.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑