Retrospective Machine Learning Exploration of Critical Factors Influencing Quality Release Test

This project was a client-based project for a biopharmaceutical company. There was a need to evaluate new growth-related parameters for predicting product quality changes. This study applied predictive analytics using ML modeling to evaluate growth-related variables and their influence on various quality release test results. The results of this analysis will be used to help support better-informed recommendations for changing process parameters before quality issues arise. Additionally, these models could help save costs associated with investigational work and lost batches from unexpected poor quality by increasing the prediction power of the current process control system. Both supervised and unsupervised modeling methods were assessed in this paper. Adopting new ML models for process control like ANN and KNN could increase product consistency and improve quality. Consequently, sustaining good product quality will also result in higher yields and increased company profitability.