论文部分内容阅读
目的应用随机森林模型法评估研究入境国际航行船舶携带输入外来病媒生物的风险。方法以中国第2大港、世界第5大港的宁波港作为研究范围,以2014年到港的国际航行船舶为研究对象,对6 051艘次船舶的33项指标展开调查,采集数据信息。对数据进行清洗及变量筛选后应用R语言编程实现随机森林模型法建模训练,并以所建模型预测新到港的1 333艘次船舶外来病媒生物携带风险。结果经过变量筛选,船舶总吨、净吨、船龄、货物种类等8个变量对入境国际航行船舶上是否可能携带外来病媒生物的风险预测有重要意义;最优模型是决策树节点变量数=3,决策树数量=500的随机森林模型;通过该模型预测船舶携带外来病媒生物风险与实际检疫结果的符合率达到81.17%,预测效果良好。结论针对高度不确定的非线性系统,应用随机森林模型法可实现更加精确的预测功能,为国境卫生检疫风险评估及预警方面的研究及应用提供理论基础。
Objective To evaluate the risk of carrying imported exotic vectors by ships entering international voyages using the random forest model method. Method Taking Ningbo Port, the second largest port in China and the fifth largest port in the world as the research area, the international voyages to Hong Kong in 2014 are investigated. 33 indicators of 6,051 vessels are surveyed to collect data and information. After cleaning the data and screening variables, the R forest programming method was used to train the stochastic forest model method and the risk of carrying foreign vectors by the newly arrived port was predicted by the model. Results After screening the variables, it is of great significance to estimate whether the eight variables, such as gross tonnage, net tonnage, age of the vessel and type of cargo, are likely to carry the risk of foreign carriers on the international voyages. The optimal model is the number of nodes in the decision tree = 3, and the number of decision trees = 500. According to the model, the coincidence rate of the biological risk of carrying foreign vectors carried by ships with the actual quarantine results reaches 81.17% and the forecasting effect is good. Conclusions For highly uncertain nonlinear systems, the application of stochastic forest model method can achieve more accurate prediction functions and provide a theoretical basis for the research and application of frontier health quarantine risk assessment and early warning.