A1380
Title: Support vector machine analysis under right censored data
Authors: Jin-Jian Hsieh - Department of Mathematics, National Chung Cheng University (Taiwan) [presenting]
Abstract: Predicting patient survival time is crucial in medical treatment, providing valuable insights for doctors and patients while guiding future treatment planning. A novel approach to applying support vector machines to right-censored data, a common challenge in survival analysis, is evaluated. The goal is to enhance support vector machines' prediction accuracy for right-censored data. To achieve this, partial rank estimation and generalized additive models are integrated to impute censored values, followed by support vector regression for model training and survival time prediction. Through simulations, the method is compared with the widely used "survivalsvm" and the Buckley-James imputation approach. Finally, the approach is validated using three real-world datasets.
