RaqForum 43 No.
375 View  • 
     ...

AUC，GINI，KS

 

AUC, GINI, KS are usually used to evaluate the overall performance of the model

Calculate AUC，GINI，KS on the prediction data “titanic_export.csv”. The variables used in the calculation are “Survived_1_percentage” and “Survived”

SPL code:

A

B

1

=T("D://titanic_export.csv")

2

=A1.(Survived_1_percentage).ranks()

3

=A1.derive(A2(#):rank)

4

=A3.groups(Survived;sum(rank):sum_rank,count(~):count)

5

=(A4(2).sum_rank-A4(2).count*(1+A4(2).count)/2)/(A4(2).count*A4(1).count)

/auc

6

=2*A5-1

/gini

7

=A1.sort@z(Survived_1_percentage)

8

=A7.len()\10+1

9

=A7.derive(#\A8:decile)

10

=A9.groups(decile;count(Survived==1):event,count(Survived==0):non_event)

11

=A10.derive(event+cum_event[-1]:cum_event,non_event+cum_non_event[-1]:cum_non_event)

12

=A11.derive(cum_event/A4(2).count-cum_non_event/A4(1).count:ks)

13

=A12.max(ks)

/ks

A5 Return AUC

..

A6 Return GINI

..

A13 Return KS

..

The calculation method of these three indicators is more complex, interested readers please consult the relevant information, this book only provides calculation code for readers to use.

About the using method of each index can see the model evaluation section of this course

Data Mining Course (raqsoft.com)


 
 •  375 View  •