Chinese Journal of Nephrology, Dialysis & Transplantation
Previous Articles Next Articles
Online:
Published:
Abstract: Objective:To investigate risk factors of renal damage in children with HenochSchnlein purpura (HSP),and to establish a prediction model for children with purpura nephritis.Methodology:533 HSP patients from January 2016 to June 2018 were selected to obtain demographic characteristics,clinical symptoms and laboratory indicators. After feature selection of 31 indicators,Logistic regression and XGBoost algorithm were used for classification prediction,and 50% crossvalidation was used to test the accuracy of algorithm. Finally,we compare the performance of two models.Results:The accuracy of XGBoost prediction model was 0818,higher than that of Logistic regression,which was 0727. The precision rate,recall rate and Fscore of XGBoost model were 0830,0930 and 0877,respectively,and the area under ROC curve was 088,each of which was higher than Logistic regression. After single factor test and ranking of important characteristics of XGBoost model,the variables in the top 10 were:antistreptococcal hemolysin ‘O’ (ASO),urinary NAG,urinary RB protein,serum IgA,age,purpura recurrence,skin purpura site,abdominal symptoms,24hour urinary protein quantification,and neutrophil percentage.Conclusion:XGBoost model can be used to predict occurrence of renal damage in HenochSchnlein purpura. Compared with the traditional Logistic regression algorithm,the prediction accuracy of XGBoost is higher.
YE Yuan,SUN Tao,SHEN Si. Prediction of renal damage in children with HenochSchnlein purpura based on machine learning[J]. Chinese Journal of Nephrology, Dialysis & Transplantation, DOI: 10.3969/j.issn.1006-298X.2020.06.005.
0 / / Recommend
Add to citation manager EndNote|Ris|BibTeX
URL: http://www.njcndt.com/EN/10.3969/j.issn.1006-298X.2020.06.005
http://www.njcndt.com/EN/Y2020/V29/I6/526