topic Re: Scoring a new dataset with Clusters created in Fastclus in New SAS User

Scoring a new dataset with Clusters created in Fastclus

MoniqueParmar — Thu, 16 May 2019 14:00:13 GMT

Hi, I have built 8 clusters using proc fastclus and I now want to apply it to a new dataset - how do I do this?

Is there something similar like proc score that is used in logistic regression?

Re: Scoring a new dataset with Clusters created in Fastclus

Rick_SAS — Thu, 16 May 2019 14:28:56 GMT

I think this thread explains the syntax that you need to use:

https://communities.sas.com/t5/SAS-Programming/Scoring-Codes-after-Proc-Cluster-rules-found/td-p/265771

Re: Scoring a new dataset with Clusters created in Fastclus

MoniqueParmar — Thu, 16 May 2019 15:01:11 GMT

Thank you

Tried to apply this and the clusters are completely different

is there another method to use similar to proc score?

How else would you be able to score customers with a cluster once you have built a segmentation?

Re: Scoring a new dataset with Clusters created in Fastclus

Rick_SAS — Thu, 16 May 2019 15:24:39 GMT

Please post the code that you tried to use.

Re: Scoring a new dataset with Clusters created in Fastclus

MoniqueParmar — Thu, 16 May 2019 15:40:21 GMT

/* original dataset and cluster */

PROC FASTCLUS
DATA=WORK.SEGBASEMIG
OUT=WORK.SEGBASEMIGTEST
outstat= OUTCLUS
MAXCLUSTERS = 8
MAXITER=1000
DELETE= 100
;
VAR

TICKETS_2_ORMORE
PARTNER_FLAG
/*PARTNER_DOMINANT*/
/*DIRECT_ONLINE_FLAG*/
DIRECT_ONLINE_DOMINANT
/*ONTHEDAY_FLAG*/
ONTHEDAY_2_ORMORE
LONG_JOURNEY_2_ORMORE
/*SINGLE_DOMINANT_FLAG*/
RETURN_DOMINANT_FLAG
OPENRETURN_DOMINANT_FLAG
/*WEEKPEAK_FLAG*/
WEEKPEAK_DOMINANT
WEEKOFFPEAK_DOMINANT
/*WEEKEND_DOMINANT*/
/*WEEKORMORE_FLAG*/
WEEKORMORE_DOMINANT
/*SAMEDAY_RETURN_FLAG*/
/*ANCIL_FLAG*/
ANCIL_2_ORMORE
TYPES_2_ORMORE
COUPLE_FLAG
/*COUPLE_2_ORMORE*/
FAMILY_FLAG
/*FAMILY_2_ORMORE*/
GROUP_FLAG
/*GROUP_2_ORMORE*/
/*SOLO_DOMINANT_FLAG*/
ROUTES_2_ORMORE
EVENT_FLAG
/*CATEGORIES_2_ORMORE*/
FREQ_JOURNEY_FLAG
/*UNI_STOP_FLAG*/
/*UNI_CITY_FLAG*/
/*UNI_PROMO_FLAG*/
/*AIRPORT_FLAG*/
AIRPORT_DOMINANT
HIGHVALUE_FLAG

;
RUN;

/* new dataset and cluster */

PROC FASTCLUS
DATA=WORK.SEGBASEMIG2
OUT=WORK.SEGBASEMIGTEST2
INSTAT= OUTCLUS
MAXCLUSTERS = 8
MAXITER=1000
DELETE= 100
;
VAR

TICKETS_2_ORMORE
PARTNER_FLAG
/*PARTNER_DOMINANT*/
/*DIRECT_ONLINE_FLAG*/
DIRECT_ONLINE_DOMINANT
/*ONTHEDAY_FLAG*/
ONTHEDAY_2_ORMORE
LONG_JOURNEY_2_ORMORE
/*SINGLE_DOMINANT_FLAG*/
RETURN_DOMINANT_FLAG
OPENRETURN_DOMINANT_FLAG
/*WEEKPEAK_FLAG*/
WEEKPEAK_DOMINANT
WEEKOFFPEAK_DOMINANT
/*WEEKEND_DOMINANT*/
/*WEEKORMORE_FLAG*/
WEEKORMORE_DOMINANT
/*SAMEDAY_RETURN_FLAG*/
/*ANCIL_FLAG*/
ANCIL_2_ORMORE
TYPES_2_ORMORE
COUPLE_FLAG
/*COUPLE_2_ORMORE*/
FAMILY_FLAG
/*FAMILY_2_ORMORE*/
GROUP_FLAG
/*GROUP_2_ORMORE*/
/*SOLO_DOMINANT_FLAG*/
ROUTES_2_ORMORE
EVENT_FLAG
/*CATEGORIES_2_ORMORE*/
FREQ_JOURNEY_FLAG
/*UNI_STOP_FLAG*/
/*UNI_CITY_FLAG*/
/*UNI_PROMO_FLAG*/
/*AIRPORT_FLAG*/
AIRPORT_DOMINANT
HIGHVALUE_FLAG

;
RUN;

Re: Scoring a new dataset with Clusters created in Fastclus

ballardw — Thu, 16 May 2019 22:48:55 GMT

Try running the second FASTCLUS code without the options MAXCLUSTERS MAXITER or DELETE