<p>Plastic injection molding (IM) is a typical multiple-input multiple-output (MIMO) complex manufacturing process widely used in modern industrial production. Determining the critical process parameters and establishing the MIMO quantitative relationship between them and the plastic product quality are two fundamental problems in IM process decision analytics. Focusing on high-dimensional process parameters and multidimensional quality indicators in the IM process, this study developed a machine learning-assisted surrogate model that integrated joint mutual information (JMI) and multi-output support vector regression (MSVR). Firstly, a JMI-based sequential search algorithm was developed to measure the association relationship between each process parameter and a multidimensional quality indicator set, and automatically select the critical process parameters of the IM process. It can effectively filter redundant information from raw industrial datasets and provide essential input features for the development of surrogate models. The MSVR model was then developed to capture the MIMO quantitative relationship between the selected critical process parameters and multidimensional quality indicator set. The proposed method can preserve complete independent variable information and avoid losing the relevance of data during training. Finally, the effectiveness of the model was verified using a real-world IM process dataset.</p>