Multimodal selective state space model-based time series classification for electricity theft detection

Wanghu Chen, Long Li, Jing Li

Abstract: Highlights•Incorporating multimodal features to smooth local fluctuations in non-stationary time series.•A Mamba-based MultiModal architecture with parameter-efficient sub-channel splitting.•Channel-wise attention and multi-scale fusion enabling long-range dependency learning.•A lightweight multimodal approach for time series classification with high adaptability.•Evaluation on 128 UCR datasets verified model’s advantages.