基于机器学习的老年人高脂血症风险预测.

Objective To explore; th(i preeliertive; value; of machine learning methods for hyperlipidemia in the e;lele;rly, and analyze; the; risk factors for hyperlipidemia in the; clelcrly. Methods The; stratified random edustcr sampling meitheiel was uscel te» se;le;e;t 5 759 edelcrly people; aged 60 anel above; in 27 communities in 9 e'ountics anel municipalities of Weifang City, anel the; health examination data of these; edelcrly people; in 2020 was collected. Missing or unusual data were; excluelcel, anel 4 534 edeler adults were; eventually inedueicel in the; stuely. Ten important risk factors we;re; scree;ne;el out by univariate; analysis, which was used to construct support vce;tor machines, decision trees, XGBoost, CATBoost anel LightGBM risk preeliertion moeleds, anel AUG (Area unelcr ROC curve), accuracy, accuraeiy, recall, anel Fl value; were; uscel te) evaluate; its performance. Results The; comprehensive; performance; e»f the; CATBoost machine; learning rnoele;! was the; best anel its AUC, precision, accuraey, recall anel Fl values were; 0. 82, 76. 49%, 92. 46%, 78. 68%, anel 0. 85, respectively. In the; process of CATBoost rnoele;! prediesting the; risk e»f hypcrlipielcmia in the; edelcrly, nine; variables were; more; important for the; pre;elie;tion e»f hyperlipidemia risk, anel their importance; was in the; orde;r of high te; bottom; systedie; blooel pressure;, waist circumference, fasting blooel glucose, boely mass inelex, diastolic blooel pressure;, smoking, age;, alcohol consumptiem, anel gender. Moreover, after rnoele;! optimization, it was found that only including these; nine; factors coulel be;tte;r pre;elie;t risks. Conclusion The; risk preeliertion mode;! of hypcrlipielcmia in the; clelcrly based on the; CATBoost machine; learning me;thod has a gejeiel prediction e;ffe;ct, which can provide; a reference; for the; pre;ve;ntie)n anel treatment of hypcrlipielcmia in the; clelcrly. [ABSTRACT FROM AUTHOR]