In estimation of multiplicative or accelerated failure time model, relative error has been recognized as another useful criterion besides the squared and absolute error criterion. The general relative error criterion introduced by Chen et al. (2013) is a unified framework for efficient estimation, which includes the least absolute relative error and least product relative error estimations as special cases. In this thesis, by combining the empirical likelihood and general relative error criterion in multiplicative models, we develop a new empirical likelihood method for inference of the unknown parameters under high-dimensional setting. The limiting theory is established for the proposed empirical likelihood statistics. We also conduct simulation studies and real data analysis to evaluate the effectiveness of the proposed method.