Search papers, labs, and topics across Lattice.
This study introduces the RMISC corpus, a large-scale, high-quality real-world multivariate time series dataset comprising around 200 datasets and 142 billion time points, aimed at enhancing the training of time series foundation models (TSFMs). By pretraining four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data, the research evaluates their zero-shot generalization capabilities across various benchmarks. The findings reveal that TSFMs trained on real-world data significantly outperform those trained on synthetic data, underscoring the importance of real-world complexity in model training.
Real-world multivariate data boosts zero-shot generalization in time series models, outperforming synthetic counterparts by a significant margin.
Recent years have witnessed the emergence of multivariate modeling using time series foundation models (TSFMs), which achieve advanced zero-shot generalization. Modern multivariate TSFMs are predominantly pretrained on multivariate synthetic data, which is easier to scale but may fail to capture the complex temporal dynamics and cross-variable relationships present in real-world time series. This raises a key question: Whether and to what extent the leading TSFMs trained with the real-world corpus perform better than those trained with synthetic data? To answer this, we establish the RMISC corpus, a considerably large-scale, high-quality, openly accessible, real-world, and multivariate time series archive that contains around 200 datasets and 142 billion time points across diverse domains. Furthermore, we pretrain four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data and evaluate their zero-shot generalization capabilities on standard in-distribution and out-of-distribution benchmarks. Experimental results show that incorporating real-world multivariate data predominantly improves the generalization performance for both univariate and multivariate TSFMs. These results provide a deeper understanding of how real-world multivariate data contributes to the development of stronger TSFMs.
