We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Erdem B. & Bouke H. - Enhancing Quality Control Efficiency: A Dynamic Risk Threshold Approach
Enhance quality control efficiency using a dynamic risk threshold approach, adjusting in real-time with pyspark, hyperopt, and MLflow, for improved recall and real-time risk scoring and alerting.
- Dynamic risk threshold adjusts itself in real-time based on the influx of supply and remaining capacity.
- Built a risk model that predicts risk scores for each supply item using pyspark, hyperopt, and MLflow.
- Implemented a random selection model to select a subset of supply items for quality control, with 2.5% initial capacity.
- Adjusted the threshold based on real-time data to ensure that the right number of items are sent for quality control.
- Achieved a recall of 8.5% for the random selection model and 4% for the baseline model.
- Created a dashboard to visualize the risk distribution, including risk scores, and used Grafana for alerting purposes.
- Scheduled daily model retraining and hyperparameter tuning using Hyperopt.
- Deployed the model to a Databricks serving endpoint for real-time inference.
- Integrated with Cerberus, a application that sends requests to the model and receives risk scores.
- Implemented a feedback loop to monitor the model’s performance and make adjustments as needed.