Sparse assortment personalization in high dimensions

Jingyu Shao, Ruipeng Dong ✉, and Zemin Zheng

International Institute of Finance, School of Management, University of Science and Technology of China, Hefei 230026, China

Graphical abstract

A penalized likelihood approach is proposed, in which the low-rank and sparsity structure are considered simultaneously.New algorithm sparse factored gradient descent (SFGD) is proposed to estimate the parameter matrix.

Public summary

■ The data-driven conditional multinomial logit choice model with customer features has a good performance in assortment personalization problem when a low-rank structure of parameter matrix is considered.

■ Our proposed method considers both low-rank and sparsity structure, which can further reduce model complexity and improve estimation and prediction accuracy.

■ New algorithm sparse factored gradient descent (SFGD) is proposed to estimate the parameter matrix, which enjoys high interpretability and efficient performance in computing.