A Comparative Study Of Prognostic Prediction Models For Small Cell Lung Cancer Patients Treated With Platinum Drugs

Small cell lung cancer(SCLC)is one of the common types of lung cancer,accounting for 10%-15% of all lung cancers.The current treatment method for SCLC patients is platinum-based drug combination chemotherapy.Due to the large difference in treatment effects between different patients,the prognosis of SCLC patients has become very important.This paper establishes a prognostic prediction model for SCLC patients through traditional statistical models and classification models in machine learning algorithms.Compare the generalization performance of different models to find a better performance prediction model for the prognosis of SCLC patients,and provide a tool for clinical treatment to assist decision-making.Search the medical records database of the First Affiliated Hospital of University of Science and Technology of China,collect patient information according to the established inclusion criteria,determine the initial variable set of prognostic factors combined with relevant medical knowledge and professional suggestions,and then use the stepwise discriminant method to screen variables,and finally get the variables that can be used for modeling and analysis.Since the data used for modeling is an unbalanced data set,it is necessary to resample the data before modeling to ensure that the number of patients with effective and ineffective treatments is consistent during modeling.Use the distance discriminant analysis of traditional discriminant analysis,the CART decision tree in machine learning,and the Bagging-DT and Bagging-Bayes model combined with Bagging ideas in ensemble learning,and the model constructed by the 10-fold cross-validation method and some evaluation indicators Perform generalization performance evaluation.Finally,the AUC value of the distance discriminant analysis is 0.6423,the AUC value of the Bagging-Bayes model is 0.603,and the prediction effects of the two models are not ideal.The CATR decision tree model in machine learning and the Bagging-DT model of integrated learning has high accuracy in predicting the prognosis of SCLC patients.The AUC value of the CART decision tree is 0.833,and the AUC value of the Bagging-DT model is 0.909.From the results,the use of machine learning decision tree algorithms to construct a prognostic prediction model for SCLC patients has certain significance for medical professionals to make treatment decisions.