使用分群方法分析登革熱資料

Analysis of dengue data with clustering

張芮瑄(Jui-Hsuan Chang)

指導教授 ： 歐陽彥正

國立臺灣大學/共同教育中心/統計碩士學位學程/碩士(2019年)

https://doi.org/10.6342/NTU201901676

摘要

2015年於台灣台南市所爆發的登革熱疫情，主要流行的登革熱為新血清型的第二型病毒，市民多對此無免疫力。透過發現不同群住院病患的特徵可以協助醫生使得患者提早發現嚴重性，並藉此降低死亡率。本研究提出了一個創新的分群方法，能框出主要的群集核心，利用住院病人之血液檢測將其分群，並找到在不同群下適當的重症指標，使其有更高的勝算比。結果顯示，兩群住院病人的主要特徵一是有明顯高燒症狀但血小板值正常，二是無明顯發燒症狀但血小板值過低。第一群人中白血球指標大於6.81時，有3.58的勝算比，第二群人中白血球指標大於7.65時，有11.18的勝算比。總體而言，KDE分群後的結果發現第二群裡之患者更為嚴重。當該組病人使用較高的WBC作為閾值時，勝算比較分群前大大提高。

關鍵字

登革熱 ； 核密度估計 ； 集群分群

並列摘要

In 2015, due to a new serotype dengue fever outbreak in Tainan City, the public was not immunized from it. Discovering the characteristics of different groups in hospitalized patients can help them to identify the high-risk groups timely and thereby reduce the case fatality rate. In this study, a novel KDE clustering was proposed to extract the main cluster cores and was utilized it to cluster inpatient into two groups by their blood tests. The results show that one group of inpatients has obviously high fever with normal platelet value, while the other group have no obvious fever but significant low platelet value. Furthermore, this study shows that if the white blood cell in the first group of people is greater than 6.81, there is an odds ratio of 3.58 for severity. If the white blood cell in the second group is greater than 7.65, there is an odds ratio of 11.18 for severity. Compare with the odds ratio of 4.8 before clustering, the results after KDE clustering find a more serious second group of patients. When this group uses a higher WBC as cutoff value, the odds ratio is greatly improved, and the results created by KDE clustering have highest odds ratio than other conventional clustering methods.

並列關鍵字

Dengue fever ； Kernel density estimation ； clustering analysis

參考文獻

[1] MacQueen, J. B., (1967). “Some methods for classification and analysis of multivariate observations,” Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, University of California Press, Vol. 1, pp.281-297.

[2] Rousseeuw, P.J., (1987). “Silhouettes: A Graphical Aid to the Interpretation and Validation of Cluster Analysis.” Journal of Computational Applied Mathematics, Vol. 20(1), pp.53–65.

[3] Kaufman, L., Rousseeuw, P.J., (1990). Finding Groups in Data: An Introduction to Cluster Analysis. John Wiley & Sons, New York.

[4] Lin, T.I., Lee, J.C., Hsieh, W.J., (2007). “Robust Mixture Modeling Using the Skew t Distribution.” Statistics and Computing, Vol. 17(2), pp.81–92.

[5] Azzalini, A., Menardi, G., (2014). pdfCluster: Cluster Analysis via Nonparametric Den- sity Estimation. R package version 1.0-2, http://CRAN.R-project.org/package= pdfCluster.