Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chpt 8- Visualizing Model Performance (ROC (X=Fase positive (False alarm),…
Chpt 8- Visualizing Model Performance
Profit curves
x= % of test instances (decreases by score)
Y= Profit
Evaluates profit at each different number of instances or rows
Uses
Eval which model can bring max profit
Eval which model is most profitable at budgeted amount of instances
ROC
X=Fase positive (False alarm)
Y= True positive (hit)
Uses confusion matrix
Data must be above diagonal to be better than chance
Stepwise starting with most highest rank
Area under curve is good one number summary
Cumulative response curve
X= % of test instances (decreasing by score)
Y= % of positives
Use
Shows how much better your model is over chance
Lift curve
A distribution of difference between chance and model
Ranking instead of classifying
Use
To have a continuous data set instead of binary