In the course of a research project with the topic “Inline Quality Control in Injection Molding” a fully interconnected manufacturing cell was built. In the experimental setup a robot is used to remove the molded parts out of the mold and for determining their weights. The measured weight and a variety of other time dependent sensor data were registered shot by shot. The goal of this thesis is to implement an interface using the programming language Python in order to automatically extract features out of this sensor data. These features were further analyzed by using a random forest to identify the most important values for quality to be able to use these values to create a statistical model for the prognosis of the part weight. With the intention of generating a broad dataset, experiments with a central composite design and consideration of five and seven different factors were made. In one case the packing pressure, the packing time, the mold temperature, melt temperature and the injection speed were examined. And in the other case the back pressure and residual cooling time were tested additionally to the other five factors. Finally linear, bilinear and quadratic statistical models with the use of a different number of influencing factors were created and their prediction accuracy was determined and compared. Additionally, it was investigated if the utilization of in-mold sensors can lead to an improvement of the model accuracy and how big the expected deviation between the predicted and the actual part weight is.