本文以气候因素及人口密度为预测变量，以系统抽样的方式选择未爆发的地区与已知的登革热爆发地区组成因变量，使用机械学习中的增强回归树模型，找出对该疾病流行影响最大的因素。结果显示影响登革热爆发的最重要因素是水汽压和温度；不同伪安全区域数据的选择方法确实会对模型拟合产生影响；基于循证共识数据的预测可能是对真实风险图的更好估计，因为它们可以消除由非报告偏差和近距离选择造成的潜在气候偏差影响。

关键词：登革热、增强回归树模型、气候因素、循证共识数据

Objective of this paper is to use evidence-based consensus data to construct a spatial distribution model of global dengue fever, to identify the important factors affecting the epidemic of dengue fever, and to examine the impact of different geographical distance selection on the model results. In this study, climatic factors and population density were used as predictors. The unexploded areas were selected by systematic sampling, and the dependent outbreaks were used to form dependent variables. Boosted regression trees in mechanical learning were used to find out the disease's greatest impact on the epidemic. Results show that the most important factors affecting the outbreak of dengue are water vapor pressure and temperature; different PA selection methods do have an impact on model fitting; ECS-based predictions might be a better estimate of real risk maps because they are based on pseudo-safety regional consensus data to eliminate the effects of non-reported bias and avoid close-range selection that could lead to climate bias.

Keywords: Dengue Fever, Boosted Regression Trees Model, Climatic Factors, Evidence-based Consensus Data