Time-varying Coefficient Cox Proportional Hazard Model with Deep Neural Network
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study presents deepTVC-Cox, a time-varying coefficient Cox model that uses deep neural networks to overcome the limitations of traditional survival analysis methods. The Cox proportional hazards model assumes that the regression coefficients remain constant over time, which may not be true for real-world survival data. To overcome this limitation, the proposed model integrates the varying-coefficient regression framework with deep learning techniques, resulting in a more flexible and data-driven method for survival modeling. The deepTVC-Cox model effectively captures complex high-dimensional relationships by learning time-dependent covariate effects using a deep neural network. This method allows the model to accommodate non-proportional hazards and improves predictive performance. Extensive numerical studies on synthetic and real-world survival datasets revealed that deepTVC-Cox outperformed the conventional Cox model, particularly when the proportional hazard assumption was violated. These findings suggest that deepTVC-Cox is a viable alternative for survival analyses in high-dimensional and time-dependent settings.
Abstract
This study presents deepTVC-Cox, a time-varying coefficient Cox model that uses deep neural networks to overcome the limitations of traditional survival analysis methods. The Cox proportional hazards model assumes that the regression coefficients remain constant over time, which may not be true for real-world survival data. To overcome this limitation, the proposed model integrates the varying-coefficient regression framework with deep learning techniques, resulting in a more flexible and data-driven method for survival modeling. The deepTVC-Cox model effectively captures complex high-dimensional relationships by learning time-dependent covariate effects using a deep neural network. This method allows the model to accommodate non-proportional hazards and improves predictive performance. Extensive numerical studies on synthetic and real-world survival datasets revealed that deepTVC-Cox outperformed the conventional Cox model, particularly when the proportional hazard assumption was violated. These findings suggest that deepTVC-Cox is a viable alternative for survival analyses in high-dimensional and time-dependent settings.