WL03: Challenges in Predictive Model Building

Music City Center 

This roundtable will focus on key challenges in predictive model building, covering feature selection, model selection and validation. We'll explore how to balance misclassification error, model sensitivity and specificity, especially in case of data with high class imbalance. Key discussion questions include: a) How can we minimize bias in training data? And b), What are effective strategies for balancing model performance? The session will foster collaboration among statisticians, data scientists and professionals, offering insights on practical strategies for building robust, responsible models. Participants will leave with actionable takeaways on creating predictive models with high precision in their fields.

Predictive Modelling

Class Imbalance

Random Forest

Logistic Regression

Neural Networks

Misclassification 

Section on Statistical Consulting