Spot potential bias during training with Qlik Predict

Qlik Predict now surfaces bias signals during model training so you can identify data and feature patterns that may lead to unfair or unreliable predictions, before you deploy your model.

Choose which features to evaluate, then review clear flags and metrics directly in the training summary. Insights include signals such as imbalanced groups and potential proxy features, helping you validate whether a feature may be acting as a stand-in for sensitive attributes.

Qlik Predict bias detection does not automatically modify your data, features, or model. It only reports detected signals, so you can decide what action—if any—to take before deployment.

Detecting bias in machine learning models