Production logging testing (PLT) is a series of monitoring programs used to evaluate multiphase fluid systems. This method can be used to determine the production profile along the horizontal wellbore section. However, traditional PLT methods are time-consuming and costly. In addition, in horizontal wells with open hole completion, it is usually difficult to apply due to the problem of wellbore collapse. In this study, a horizontal well PLT analysis and prediction method based on the GRU is established. Firstly, static and dynamic data are collected, including permeability, porosity, depth, oil/gas/water production, etc., outlier analysis is performed based on the box principle, Kalman denoising method is used to achieve an information retention of 98%; Secondly, based on the physical characteristics of the reservoir through which the horizontal well passes, the target well is divided into 10 sections, and the dynamic weights of each section on the total production capacity are defined and calculated; Thirdly, all sections are input into multi-mode algorithms GRU, then the production capacity of each well section are trained and predicted; Finally, compared with reservoir numerical simulation (RNS), this method reduces error by 45% and improves accuracy by 32%, effectively achieving production prediction of the target reservoir. This method has been successfully applied to an open hole completion. The method establishes a workflow for dynamic splitting of horizontal wells and achieves high-precision dynamic prediction based on AI, providing reference for reservoir engineers in similar reservoirs to make decisions under the same conditions.