Application of different machine learning algorithms in the construction of prognostic models for rectal squamous cell carcinoma:based on SEER database

HUANG Shi; LIN Dandan

doi:10.12173/j.issn.1004-5511.202410124

Abstract：Objective To investigate the application value of different machine learning models in predicting 5-year survival in patients with rectal squamous cell carcinoma(rSCC).Methods Data from patients diagnosed with rSCC between 2004 and 2015 were collected using SEERStat software in SEER database.Patients were randomly divided into a training set and a validation set in a 7∶3 ratio.Models were constructed using extreme gradient boosting(XGBoost),random forest,support vector machine,and k-nearest neighbor algorithms on the training set.The predictive ability of the models was evaluated using the area under the receiver operating characteristic curve(AUC),calibration curves,and decision curves.The SHAP algorithm was used to identify the contribution of variables to the model for the best-performing model.Results A total of 833 patients with rSCC were included,including 584 in the training set and 249 in the validation set.Multiple machine learning models were constructed based on 10 variables:age,sex,race,marital status,T stage,N stage,M stage,surgery,chemotherapy,and radiotherapy.In the validation set,the XGBoost model performed best[AUC=0.758,95％CI(0.696,0.820)],demonstrating moderate predictive ability and good calibration.The decision curves demonstrated high clinical value.SHAP analysis showed that T stage contributed most to the XGBoost model's decision-making,while radiotherapy contributed least.Conclusion This study constructed an interpretable XGBoost model that can assist physicians in assessing the 5-year survival and treatment efficacy of rSCC patients and in developing personalized treatment plans.