Metal–organic chemical vapor deposition (MOCVD) is a vapor-phase epitaxial growth technique commonly used in industry and academia to manufacture high-quality Gallium Arsenide (GaAs) compound semiconductor materials. In the actual industrial GaAs MOCVD epitaxial production process, a large amount of process parameters, sensor data, epitaxial wafer testing data, etc. are generated. These epitaxial-growth big data have large sample size and high noise, making data mining increasingly difficult. Nevertheless, this indicates that these data contain abundant valuable information. Therefore, this study proposes a machine learning–based GaAs MOCVD growth model. By analyzing the GaAs epitaxial growth data generated during MOCVD of GaAs, a machine learning model is established to predict the optimal production process for epitaxial growth. By mining the relationship between variables and developing a predictive model for assessing the performance of epitaxial materials, important guidance can be provided for process parameter tuning of GaAs epitaxial growth, considerably reducing debugging costs and further improving the intelligence level of GaAs MOCVD epitaxial growth.