关键词: 时间序列, 表示学习, 滤波器, 扩张卷积, 自监督, 时间序列预测

Abstract: Time series data is widely used in fields such as power load forecasting and meteorological analysis. Extracting high-quality representations is crucial for downstream prediction tasks. However, the performance of existing methods is limited by high-frequency noise interference, difficulty in modeling long-term dependencies, and the scarcity of labels. Therefore, a time series representation method based on Spectral Filtering and Hierarchical Dilation (SFHD) was proposed. Firstly, a Spectral Filtering Block (SFB) was designed to extract multi-scale features through global and local filters, and an adaptive spectral filtering mechanism was used in the frequency domain to weaken the influence of high-frequency noise. Then, a Hierarchical Dilation Block (HDB) was constructed, in which exponentially dilated convolutions were utilized to obtain a progressively enlarged receptive field, thereby enhancing the ability to capture long-term dependencies. Finally, a change-aware self-supervised pretraining strategy was proposed, in which highly dynamic segments were masked to force the model to learn the underlying structure of the sequence, thereby alleviating the problem of insufficient labeled data. Experimental results on seven public datasets demonstrate that, compared with the suboptimal model iTransformer, SFHD achieves an average reduction of 9.47% in Mean Square Error (MSE) and 5.36% in Mean Absolute Error (MAE). These results indicate that SFHD provides stronger representation capabilities and leads to improved performance on downstream time series prediction tasks.

Key words: time series, representation learning, filters, dilated convolution, self-supervision, time series prediction