Objective To analyze the risk factors of cerebral arterial vessel wall calcification in ca-rotid atherosclerosis(CAS)patients by applying random forest model and logistic regression model.Methods A total of 180 CAS patients admitted to Department of Neurosurgery of Heng-shui People's Hospital from August 2021 to February 2024 were enrolled,and divided into a train-ing set(126 patients)and a testing set(54 patients)in a ratio of 7∶3.The patients in the training set were divided into the calcification group(42 cases)and the non-calcification group(84 cases)according to the results of CTA for calcification in the cerebral arterial vessel wall.General clinical data were compared between the two groups to screen the variables with statistical differences,and these variable were taken into the random forest model and logistic regression model.Risk factors for CAS patients with cerebral arterial wall calcification were analyzed,and the predictive performance of the 2 models was compared with ROC curve analysis.Results The results of ran-dom forest algorithm indicated that the top 5 risk factors affecting cerebral arterial wall calcifica-tion in CAS patients were hypertension,diabetes,LDL-C,age,and homocysteine.Multivariate lo-gistic regression analysis model revealed that age,hypertension,diabetes,LDL-C and homocys-teine were risk factors for cerebral arterial wall calcification in CAS patients(OR=1.039,95％CI:1.009-1.075;OR=1.006,95％CI:1.001-1.023;OR=2.053,95％CI:1.341-3.172;OR=1.687,95％CI:1.116-3.304;OR=1.149,95％CI:1.007-1.291).The prediction indicators of random forest model were better than those of logistic regression analysis model,but the differ-ence between training set and testing set was greater than that of logistic regression analysis mod-el.The stability of logistic regression analysis model was better,and the prediction efficiency of combined model was better than that of single model.Conclusion The random forest model has a higher predictive efficacy,and the logistic regression model is more stable,so that the combination of the two models has a higher predictive value.