Abstract:Objective To explore the predictive walue and effect of logistic regression and support vector machine (SVM) model in the formation of pelvic lymphocysts uithin 14 days after cervical cancer surgery. Methods A retrospective analysis was conducted on the clinical data of 128 cervical cancer patients who underwent surgicaltreatment at Northern Theater Command General Hospital from January 2020 to January 2024. The data were divided into a training set (n=90) and a test set (n=38) in a 7∶3 ratio. In the training set, univariate and multivariate logistic regression were used to analyze the risk factors for the formation of pelvic lymphocysts within 14 days after cervicalcancer surgery. Logistic regression and SVM model were constructed by using the logistic regression variable screeningstrategy. Receiver operating characteristic (ROC) curves were drawn for both models in the training and test sets to compare their predictive performance. Results The incidence of pelvic lymphocysts in cervical cancer patients 14 days after surgery was 38.89% (35/90) in training set and 42.11% (16/38) in test set. Multivariate logistic regression analysis showed that open abdominal surgery, surgical instruments using unipolar electrosurgical scalpel, number of removed lymph node > 20, pelvic and para-aortic lymphadenectomy, drainage days > 3 days, postoperative chemoradio- therapy were the risk factors for pelvic lymphocysts formation within 14 days after cervical cancer surgery (P＜0.05) . The importance of the predictive variables, in descending order, were lymph node dissection range, postoperativeradiotherapy/chemotherapy, number of lymph nodes removed, drainage duration, use of unipolar electrosurgical scalpel, and use of open abdominal surgery. The SVM model consisted of open abdominal surgery, unipolarelectrosurgical scalpel, number of removed lymph node, lymphadenectomy range, drainage days, postoperativeradiotherapy and chemotherapy. The accuracy (88.22%, 78.95%) and AUC (0.785, 0.776) of SVM model in predicting the formation of pelvic lymphocysts after cervical cancer surgery in both the training and test sets were higherthan those of logistic regression model (75.56%, 71.05%) and AUC (0.712, 0.694) . The difference of AUC between the two models was statistically significant (training set: Z=8.655,P=0.001; test set: Z=2.454,P=0.003). Conclusion The accuracy, sensitivity, specificity, AUC and other indicators of SVM model in predicting the formation of pelvic lymphocysts within 14 days after surgery in patients with cervical cancer are better than logistic regression model.