Search papers, labs, and topics across Lattice.
This paper introduces SCPaT, a Transformer-based framework for multivariate time series forecasting that utilizes semantic structured partitioning to overcome limitations of existing patch-based methods. By decomposing input sequences into semantically consistent units and constructing a dynamic semantic graph to model interactions among these units, SCPaT effectively captures complex temporal patterns. Experimental results on 12 real-world datasets show that SCPaT outperforms traditional forecasting approaches, highlighting its potential for improved accuracy in MTSF tasks.
Semantic structured partitioning in SCPaT leads to enhanced forecasting accuracy by intelligently modeling interactions among heterogeneous temporal patterns.
Multivariate time series forecasting (MTSF) is a fundamental task in many real world applications. Existing patch based forecasting methods generally fall into three categories: fixed partitioning, multi-scale partitioning, and extendable partitioning. Fixed partitioning often breaks meaningful temporal boundaries, multi-scale partitioning may introduce redundant representations across scales, and extendable partitioning improves flexibility but still lacks an explicit mechanism for organizing semantic structure and modeling interactions among heterogeneous temporal patterns. To address these limitations, we propose SCPaT, a Transformer based framework built on semantic structured partitioning. SCPaT first decomposes input sequences into semantically consistent units through adaptive semantic unit generation, then constructs a dynamic semantic graph to model directed dependencies among these units and organize them into higher order semantic blocks. Based on these structured representations, an importance aware routing mechanism adaptively dispatches different semantic blocks to different experts for customized modeling. Extensive experiments on 12 real world datasets demonstrate the effectiveness of SCPaT.
