<p>In some research scenarios, researchers may prioritize relative errors to reflect the significance of errors in relation to the size of the data. In this paper, the least product relative error (LPRE) criterion is considered for the multiplicative model with positive response variables. In the face of the situation that the target data modeling is difficult to support the subsequent analysis due to the lack of data, transfer learning is used to improve the prediction performance by using similar datasets. This paper extends the transfer learning framework to the high-dimensional multiplicative regression model and proposes a two-step transfer learning algorithm as well as a transferable source detection algorithm based on the LPRE criterion. Both the relative error and the positive response variable are concerned. We validate the method’s performance using numerical simulations and then apply it to restaurant revenue dataset.</p>