Objective: To investigate the significant factors influencing the need for perioperative blood transfusion in patients undergoing hepatectomy. Methods: Medical records of patients who underwent elective hepatectomy in our hospital from January 2020 to December 2021 were retrospectively collected. Variables associated with transfusion were analyzed using traditional logistic regression (LR) and machine learning algorithm, the Classification and Regression Tree (CRT). The predictive values of the two methods were compared by the area under the curve (AUC) of the ROC curve. Results: Among the 402 patients, 82(20.4%) received blood transfusions. Multivariable logistic regression analysis identified several risk factors for perioperative blood transfusion, including vascular invasion, preoperative hemoglobin level, intraoperative blood loss, duration of surgery, postoperative hemoglobin level, and postoperative complications (P<0.05). In the CRT model for predicting blood transfusion, intraoperative blood loss (cutoff: 450 mL) was the parent node, with preoperative Hb, postoperative complications, and hospital stay as child nodes. The LR model demonstrated superior predictive performance compared to the CRT model, with an AUC of 0.971 (95% CI: 0.956-0.985) vs 0.937 (95% CI: 0.909-0.965). The difference in AUC between the two methods was statistically significant (P<0.05). Conclusion: Although the CRT model did not outperform logistic regression in overall predictive accuracy, it still provides a valuable tool for assisting clinicians in making decisions about blood transfusion in the perioperative period of hepatectomy, thereby facilitating more individualized guidance for preoperative blood preparation in clinical practice.