Bi-level feature selection in high dimensional AFT models with applications to a genomic study

Area of Science:

• Statistics
• Bioinformatics
• Machine Learning

Background:

• High-dimensional data presents challenges in statistical modeling.
• Accelerated failure time (AFT) models are crucial for survival data analysis.
• Feature selection is vital for improving model interpretability and performance.

Purpose of the Study:

• To develop a novel bi-level feature selection method for high-dimensional AFT models.
• To formulate AFT models into a single index model for efficient feature selection.
• To provide a computationally efficient and easily implementable algorithm.

Main Methods:

• A new bi-level feature selection approach is proposed.
• The method formulates high-dimensional AFT models into a single index model.
• An expedient algorithm is developed for sparse solutions at group and individual feature levels.

Main Results:

• The proposed method achieves sparse solutions at both group and individual feature levels.
• The algorithm is computationally efficient and easy to implement.
• Demonstrated effectiveness through genomic data analysis and simulation studies.

Conclusions:

• The novel bi-level feature selection method offers an efficient solution for high-dimensional AFT models.
• The approach enhances model interpretability and performance by selecting relevant features.
• The method is validated through practical data analysis and simulation, showing good finite sample performance.