Regional VOCs Gathering Situation Intelligent Sensing Method Based on Spatial-Temporal Feature Selection

As VOCs pose a threat to human health, it is important accurately capture changes in concentrations and sense relevant areas. Therefore, necessary improve the accuracy of concentration prediction realise aggregation situation sensing. Firstly, on basis regional grid division, inverse distance spatial interpolation method used for collect data information. Secondly, extreme gradient boosting (XGBoost) spatio-temporal feature selection, combined with graph convolutional neural network (GCN) construct relationships VOCs, multiple linear regression (MLR) process time series predict grid. Finally, potential values are calculated based results, perception results visualised. A proposed, using XGBoost-GCN-MLR scenario-aware approach perceive region. trend were carried out Xi’an, Baoji, Tongchuan, Weinan Xianyang. The show that compared GCN model, XGBoost MLR model GCN-MLR reduces input variables, achieves optimisation parameters complexity improves accuracy. Intelligent sensing can visualise VOCs. intelligent development status convey more information, which has practical value.