Evaluation of machine learning algorithms for the prognosis of breast cancer from the Surveillance, Epidemiology, and End Results database

• Published in: PloS one 2023-01, Vol.18 (1), p.e0280340-e0280340
• Main Authors: Wu, Ruiyang, Luo, Jing, Wan, Hangyu, Zhang, Haiyan, Yuan, Yewei, Hu, Huihua, Feng, Jinyan, Wen, Jing, Wang, Yan, Li, Junyan, Liang, Qi, Gan, Fengjiao, Zhang, Gang
• Format: Article
• Language: English
• Subjects: Algorithms; Applications programs; Artificial intelligence; Breast cancer; Breast Neoplasms - therapy; Cancer; Chemotherapy; Cohort Studies; Comparative studies; Computer and Information Sciences; Confidence intervals; Data analysis; Data mining; Databases, Factual; Decision trees; Development and progression; Digital technology; Discriminant analysis; Epidemiology; Female; Health surveillance; Humans; Information management; Internet software; Learning algorithms; Machine Learning; Medical diagnosis; Medical prognosis; Medicine and Health Sciences; Modelling; Oncology, Experimental; Optimization techniques; Patients; Performance evaluation; Physical Sciences; Precision medicine; Predictive Value of Tests; Prognosis; Research and Analysis Methods; Spline functions; Statistical analysis; Statistics; Support vector machines; Surveillance; Survival; Survival analysis; Variables
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0280340

Many researchers used machine learning (ML) to predict the prognosis of breast cancer (BC) patients and noticed that the ML model had good individualized prediction performance. The cohort study was intended to establish a reliable data analysis model by comparing the performance of 10 common ML algorithms and the the traditional American Joint Committee on Cancer (AJCC) stage, and used this model in Web application development to provide a good individualized prediction for others. This study included 63145 BC patients from the Surveillance, Epidemiology, and End Results database. Through the performance of the 10 ML algorithms and 7th AJCC stage in the optimal test set, we found that in terms of 5-year overall survival, multivariate adaptive regression splines (MARS) had the highest area under the curve (AUC) value (0.831) and F1-score (0.608), and both sensitivity (0.737) and specificity (0.772) were relatively high. Besides, MARS showed a highest AUC value (0.831, 95%confidence interval: 0.820-0.842) in comparison to the other ML algorithms and 7th AJCC stage (all P < 0.05). MARS, the best performing model, was selected for web application development (https://w12251393.shinyapps.io/app2/). The comparative study of multiple forecasting models utilizing a large data noted that MARS based model achieved a much better performance compared to other ML algorithms and 7th AJCC stage in individualized estimation of survival of BC patients, which was very likely to be the next step towards precision medicine.