Becton Dickinson’s (BD) Medication Management Solutions (MMS) Division needed a way to classify over 500,000 complaints for their product line to meet FDA compliance standards. To solve this problem, Analytica developed a machine learning model using natural language processing which parsed and classified each complaint so that BD could pass its FDA audit.
PROJECT GOAL
- FDA audit yielded 500,000 unclassified product complaints
- All previous and future complaints must be classified using an automatic solution with 80% accuracy
PROJECT RESULTS
- Developed an unsupervised machine learning complaint reclassification model using natural language processing
- Model is currently matching at a 95% accuracy
