IoT Data Imputation Accuracy Enhancement: A Spatiotemporal Causal Mamba-Diffusion Imputation Model

The accuracy of time-series data from Internet of Things (IoT) sensors plays a crucial role in contemporary intelligent transportation systems (ITSs). However, in the process of IoT data collection, the frequent occurrence of missing data poses significant challenges for data accuracy, thereby limiting the service-provisioning of ITS. To solve this issue, this article proposes a deep-learning-based data imputation model, namely, spatiotemporal causal Mamba-diffusion imputation (SCMDI) model, which effectively models complicated nonlinear dependencies and captures the implicit causal relationships inherent in traffic patterns. Technically, the innovative SCMDI model integrates the diffusion model with the Mamba module, enhancing the ability to capture complex dynamic interactions across space and time. In SCMDI, we first propose a Mamba-attention dual module (MADM) to further improve the joint modeling of spatiotemporal features. Additionally, SCMDI incorporates spatiotemporal causal prior knowledge, which not only improves imputation accuracy by capturing intrinsic relationships in IoT data but also effectively mitigates false correlations. Extensive experiments on three publicly available traffic datasets demonstrate that our SCMDI model outperforms state-of-the-art methods, achieving a 10% improvement in data accuracy, and proving its efficiency in handling complex spatiotemporal data in IoT data imputation. The ablation experimental results further validate the effectiveness of causal prior and MADM.