Identify Influential Observations in Binary Classification
Internal function: Build an AUC ggplot2 object
Internal function: Create lift-chart ggplot2 object
Internal function: Fetch indeces in an output object
Influence Functions On AUC
Cumulative Lift Charts
Local Influence Approaches On AUC
Determine Identified Influential Cases
Visualize IAUC result
Visualize ICLC results
Visualize LAUC results
Internal function: Plot sequentially
Show IAUC results
Show LAUC results
Internal function: Print output
Ke, B. S., Chiang, A. J., & Chang, Y. C. I. (2018) <doi:10.1080/10543406.2017.1377728> provide two theoretical methods (influence function and local influence) based on the area under the receiver operating characteristic curve (AUC) to quantify the numerical impact of each observation to the overall AUC. Alternative graphical tools, cumulative lift charts, are proposed to reveal the existences and approximate locations of those influential observations through data visualization.