Section outline

  • Lecture: Model/algorithm evaluation/choice

    Model/algorithm evaluation for each data type: e.g. for binary accuracy, AUC for ROC, and Gini Index. Concept of “predictor importance”. Training-test, training-validation-test, cross-validation, nested cross-validation, reserving data -- cross-validate ~2/3 of data and reserve ~1/3 for model evaluation.

    Faculty:  John Kornak

    Location: MH-1407