基于改进随机森林算法的通榆河水生态健康状况评价研究

Study on the evaluation of water ecological health status of Tongyu River based on improved random forest algorithm

通过改进传统的随机森林算法,对江苏省通榆河的水生态健康状况进行评价研究,建立了包含18个具体指标的通榆河水生态健康状况评价体系,并将评价结果分为"病态、微病态、亚健康、微健康、健康"5个等级。模型改进结果表明,改进的随机森林算法(IRF)较传统的随机森林(RF)算法和径向基人工神经网络(ANN-RBF)在模型误差和计算效率上均有明显提升,算法性能得到较大改善;评价结果显示,通榆河在2016~2018年的水生态健康状况分别为微病态、亚健康、微健康,整体呈现良好的发展态势。改进的模型可为相关评价研究提供借鉴和指导。

The water ecological health status of Tongyu River in Jiangsu Province was evaluated by improved random forest algorithm,the evaluation system of Tongyu River water ecological health status including 18 specific indicators was established,and the evaluation results were divided into 5 grades of morbidity,micro-morbidity,sub-health,micro-health,and health.The improved model results showed that IRF had obvious improvement in model error and computational efficiency compared with RF and ANN-RBF,and the algorithm performance was greatly improved.The evaluation results showed that the water ecological health status of Tongyu River in 2016-2018 was micropathological,sub-health and micro-health,and the overall development trend was good.The improved model could provide reference and guidance for relevant evaluation research.