Evaluating and deploying predictive models
May 20 @ 3:00 pm - 4:00 pm
This webinar will explore the question of how data analysts evaluate and choose predictive models as well as the various options for deploying them into the real world. Experienced data analysts are well aware that there is no single, universal criterion for assessing the accuracy of predictive models.
This session will compare and contrast the various measures that can be employed to evaluate model performance and focus on the importance of establishing practical success criteria early in the project cycle. The webinar will also look at how model scores can be tested, and the various options for deploying the results to the wider organisation.
- What does ‘success’ look like?
- Comparing measures of predictive accuracy (e.g. overall accuracy, AUC, Lift values)
- Using charts to evaluate model performance (e.g. gains curves, ROC charts)
- Selecting models based on profitability
- Testing model performance
- Deploying models
Who should attend?
- Anyone who is building predictive models or interested in their deployment