Background Dengue fever is a globally prevalent arbovirus disease that poses a serious challenge to global health. Therefore, analyzing the relationship between dengue fever incidence and meteorological factors and developing a more effective prediction model based on this relationship can provide a theoretical basis for public health departments to formulate reasonable prevention strategies. Methods We collected dengue fever cases and meteorological data, including temperature, humidity, sunshine duration, etc., from Guangdong and Zhejiang Provinces in China from 2005–2024. A distributed lag nonlinear model (DLNM) was used to analyze the exposure–response relationship between meteorological factors and dengue incidence. Moreover, the raw case data were classified into dengue warning levels using a fuzzy clustering algorithm. The improved horned lizard optimization algorithm (IHLOA) was then combined with support vector machine (SVM), random forest (RF) and k-nearest neighbor (KNN) for dengue prediction. The average accuracy (<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq1.gif" Format="GIF" Height="18" Rendition="HTML" Resolution="72" Type="Linedraw" Width="49" /> </InlineMediaObject> <EquationSource Format="TEX">${Avg}_{acc}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">Avg</mi> </mrow> <mrow> <mi mathvariant="italic">acc</mi> </mrow> </msub> </math></EquationSource> </InlineEquation>), average fitness value (<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq2.gif" Format="GIF" Height="20" Rendition="HTML" Resolution="72" Type="Linedraw" Width="49" /> </InlineMediaObject> <EquationSource Format="TEX">${Avg}_{fit}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">Avg</mi> </mrow> <mrow> <mi mathvariant="italic">fit</mi> </mrow> </msub> </math></EquationSource> </InlineEquation>), average feature reduction rate (<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq3.gif" Format="GIF" Height="20" Rendition="HTML" Resolution="72" Type="Linedraw" Width="74" /> </InlineMediaObject> <EquationSource Format="TEX">${Avg}_{feature}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">Avg</mi> </mrow> <mrow> <mi mathvariant="italic">feature</mi> </mrow> </msub> </math></EquationSource> </InlineEquation>), standard deviation (STD) and <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq4.gif" Format="GIF" Height="18" Rendition="HTML" Resolution="72" Type="Linedraw" Width="106" /> </InlineMediaObject> <EquationSource Format="TEX">${F1\_score}_{micro}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi>F</mi> <mn>1</mn> <mi>_</mi> <mi>s</mi> <mi>c</mi> <mi>o</mi> <mi>r</mi> <mi>e</mi> </mrow> <mrow> <mi mathvariant="italic">micro</mi> </mrow> </msub> </math></EquationSource> </InlineEquation> were used to evaluate prediction performance. Results The incidence risk of dengue fever was positively correlated with temperature, relative humidity, sunshine duration and the vegetation index but negatively correlated with visibility, wind speed and sea level pressure. Meteorological factors had a lag effect on the risk of dengue fever, and the magnitude of the effect varies dynamically with lag time. Compared with the other prediction models, our proposed hybrid prediction models exhibited relatively low <InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq5.gif" Format="GIF" Height="20" Rendition="HTML" Resolution="72" Type="Linedraw" Width="74" /> </InlineMediaObject> <EquationSource Format="TEX">${Avg}_{feature}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">Avg</mi> </mrow> <mrow> <mi mathvariant="italic">feature</mi> </mrow> </msub> </math></EquationSource> </InlineEquation> values and relatively high <InlineEquation ID="IEq6"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12889_2025_22375_Article_IEq6.gif" Format="GIF" Height="18" Rendition="HTML" Resolution="72" Type="Linedraw" Width="45" /> </InlineMediaObject> <EquationSource Format="TEX">${\text{Avg}}_{\text{acc}}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mtext>Avg</mtext> <mtext>acc</mtext> </msub> </math></EquationSource> </InlineEquation> values, indicating the best prediction results. Conclusion Our experiment revealed the correlation and lag effect between meteorological factors and the incidence of dengue fever, indicating that meteorological factors have important value in predicting dengue fever. In addition, the hybrid prediction models constructed in this article can accurately predict outbreaks of dengue fever, which can lay a theoretical foundation for the construction of monitoring and early warning systems and improve the ability of relevant government departments to detect and identify dengue fever outbreaks in a timely manner.

Integrating meteorological data and hybrid intelligent models for dengue fever prediction

期刊论文
出版模式：

开放获取
发表日期：

2025年4月23日

查看全文

Yunyun Cheng,
Rong Cheng,
Ting Xu,
Xiuhui Tan,
Yanping Bai,
Jing Yang

摘要

Background

Dengue fever is a globally prevalent arbovirus disease that poses a serious challenge to global health. Therefore, analyzing the relationship between dengue fever incidence and meteorological factors and developing a more effective prediction model based on this relationship can provide a theoretical basis for public health departments to formulate reasonable prevention strategies.

Methods

We collected dengue fever cases and meteorological data, including temperature, humidity, sunshine duration, etc., from Guangdong and Zhejiang Provinces in China from 2005–2024. A distributed lag nonlinear model (DLNM) was used to analyze the exposure–response relationship between meteorological factors and dengue incidence. Moreover, the raw case data were classified into dengue warning levels using a fuzzy clustering algorithm. The improved horned lizard optimization algorithm (IHLOA) was then combined with support vector machine (SVM), random forest (RF) and k-nearest neighbor (KNN) for dengue prediction. The average accuracy ( ${Avg}_{acc}$ ${Avg}_{acc}$ ), average fitness value ( ${Avg}_{fit}$ ${Avg}_{fit}$ ), average feature reduction rate ( ${Avg}_{feature}$ ${Avg}_{feature}$ ), standard deviation (STD) and ${F1\_score}_{micro}$ ${F 1_s c o r e}_{micro}$ were used to evaluate prediction performance.

Results

The incidence risk of dengue fever was positively correlated with temperature, relative humidity, sunshine duration and the vegetation index but negatively correlated with visibility, wind speed and sea level pressure. Meteorological factors had a lag effect on the risk of dengue fever, and the magnitude of the effect varies dynamically with lag time. Compared with the other prediction models, our proposed hybrid prediction models exhibited relatively low ${Avg}_{feature}$ ${Avg}_{feature}$ values and relatively high ${\text{Avg}}_{\text{acc}}$ ${Avg}_{acc}$ values, indicating the best prediction results.

Conclusion

Our experiment revealed the correlation and lag effect between meteorological factors and the incidence of dengue fever, indicating that meteorological factors have important value in predicting dengue fever. In addition, the hybrid prediction models constructed in this article can accurately predict outbreaks of dengue fever, which can lay a theoretical foundation for the construction of monitoring and early warning systems and improve the ability of relevant government departments to detect and identify dengue fever outbreaks in a timely manner.