本项目以5G农业物联网场景下混合监测数据流驱动的作物生长状态预测作为主要研究内容。首先，在深入分析5G农业物联网混合监测数据流的特征，建立可自适应动态更新聚类簇形态的增量聚类模式，获取环境因子与作物生长状态深层关联信息，具体包括：研究混合数据流中数值型数据和属性类数据的加权相似性计算方法、基于异常点感知的概念漂移检测算法，以及基于概念漂移检测的混合数据流增量聚类算法。以此为基础，探寻环境因子与作物生长状态所对应的深层关联机理，生成序列模式。其次，研究环境因子特征信息的智能变量筛选方法，剔除与当前作物生长状态无关、弱相关的环境变量，提升预测模型泛化能力与预测精度。最后，研究面向环境因子变化趋势分析的混合数据智能组合预测方法，并利用环境因子变化趋势分析结果，匹配环境因子与作物生长状态序列模式，实现农作物生长状态预测，并评价预测结果，为5G农业物联网场景下高效农作物生产管理提供决策支持。

This project takes the crop growth state prediction driven by mixed monitoring data flow under 5G agricultural Internet of things (IoT) scenario as the main research content. Firstly, based on in-depth analyzing the characteristics of mixed data stream, the incremental clustering model, which can update clusters morphology adaptively and dynamically, will be built to obtain the deep correlation information between environmental factors and crop growth status. The specific research contents include: the weighted similarity calculation method of numerical data and attribute data in mixed data flow, the outlier perception based concept drift detection algorithm, and the concept drift detection based incremental mixed data stream clustering algorithm. On this basis, the deep correlation mechanism between environmental factors and crop growth state will be explored, and sequence association template can be generated. Secondly, the intelligent variable selection method of environmental factor feature information will be studied to eliminate the environmental variables which are not related to the current crop growth status, and to improve the generalization ability and prediction accuracy of the prediction model. Finally, the intelligent combined prediction model based on mixed data for trend analysis of environmental factors will also be studied, and the results of trend analysis of environmental factors will be used to match the sequence association template of environmental factors and crop growth status, so as to realize crop growth status prediction. The evaluation of prediction results will also be analyzed to provide efficient decision support for crop production management under 5G+IoT scenario.