Shale gas production forecasting is an important research topic. Due to the difficulties faced by conventional methods, data-driven methods are becoming increasingly popular. This study therefore presents a shale gas production forecasting model based on deep learning. The proposed model applies the multi-well scheme and does not require a long piece of production history from the target well. The proposed model also includes geological and engineering parameters to improve its accuracy. The evaluation indicates that the proposed model notably surpasses the conventional Arps decline curve and deep learning models without geological or engineering parameters. The proposed model’s overall mean absolute percentage error for cumulative gas and water production are 0.224 and 0.236, respectively.