A self-calibration model building method for rapid total nitrogen detection based on semi-supervised learning

•A model self-calibration framework is proposed, which can establish accurate TN detection model withnoising data.•The structured sparse learning ang adaptive graph learning are first unified in a semi-supervised learning paradigm.•The temporal connection of collected water samples is considered for the first time, which enhances detection performance.•Results on real-world datasets demonstrate the robustness and superiority of our approach.