Every day-increasing connectivity and access to data can provide valuable insight to the plastics industry. While the amount of accessible data has been increasing, the means to process and store it efficiently while squeezing valuable process information out of it has not been prioritized. The increase in connectivity has led to much of this data being stored and used in cloud computing systems which can be both monetarily and computationally expensive. Motivated by this fact, the feasibility of using real-time data directly captured from injection molding machine is investigated in terms of their capabilities for online quality monitoring. Using the built-in sensors that are usually existed in the standard injection molding machines (barrel pressure, screw position, and clamp force) and a dimensional reduction method, models are derived to predict quality of injection molded parts (Weight, Thickness, and Diameter). The developed models show high predictive capability with R2 values ranging from 0.89–0.97. Moreover, the combination of the proposed feature extraction method and implementation of Partial Least Squares Regression (PLS) demonstrates that most of the computing for automatic quality control can be done on local edge computing hardware with a significantly summarized data, and only control commands need to be sent through the cloud.