Abstract:[Objective]: To construct predictive models for perioperative deep vein thrombosis (DVT) risk in total knee arthroplasty (TKA) based on Logistic regression and XGBoost algorithm respectively. [Methods]: From December 2017 to October 2021, 3711 patients who underwent TKA surgery in the Department of Orthopedics, The First Affiliated Hospital of the University of Science and Technology of China were retrospectively enrolled. The logistic regression and XGBoost algorithm prediction models were constructed. Predictors and predictive power of the two were compared. [Results]:A total of 3711 patients were included in the study, including 889 in the DVT group and 2822 in the non-DVT group.Logistic regression model showed prolonged postoperative hospital stay (OR=1.393, P<0.001), advanced age (OR=1.214, P<0.001), preoperative D-dimer (OR=1.058, P=0.008), postoperative blood Increased phosphorus (OR=1.160, P =0.005) and increased postoperative urea nitrogen-to-creatinine ratio (OR=1.576, P <0.001) were risk factors for DVT events; prolonged preoperative preparation time (OR=0.854, P= 0.008) and increased preoperative prothrombin activity (OR=0.817, P=0.028) were protective factors for DVT events.Logistic regression model showed that preoperative waiting time, postoperative hospital stay, low molecular weight heparin use, Factor Xa inhibitor use, early anticoagulation interventionet al. were the predictors of perioperative DVT events in patients with TKA surgery (P<0.05). The XGBoost model showed that age, postoperative hospital stay, postoperative D-dimer, serum urea nitrogen/creatinine ratio, and use of low molecular weight heparin were important predictive feature vectors. The areas under the receiver operating characteristic curve for the two were 0.709 and 0.840, respectively.[Conclusion]: The XGBoost model has good predictive ability for DVT events in the perioperative period of TKA. Patient age, postoperative hospital stay, postoperative D-dimer, serum urea nitrogen/creatinine ratio, and use of low molecular weight heparin are potential important predictors.