9.7 Summary

Profile data has a specific structure, can be over time, have highly correlated features, and contain a hierarchical structure. Understanding the experimental unit is essential to make decisions around preprocessing and evaluation.

Basic preprocessing steps for profiled data can include reducing baseline effect, reducing noise across the profile, and harnessing the information contained in the correlation among predictors.

No one set of preprocessing steps or model will work best in every situation. Finding the right combination, along with using expert knowledge, can produce a very effective model.

Fin