Customer Classification and Consumption Habit

[summy]

Here is the sample dataset:

CustomerNo  M1T1  M1A1  M2T2  M2A2     ......   M19T19  M19A19  M20T20  M20A20

1                     2        500.4    12    3689.9                    0             0            4             12

2                     5           30       3      3000                      1        9999           9          29875

3                     0          0          0       0                           0             0            0              0

4                     1        42310    0       0                           0             0            0              0

.

.

.

.

.

.

490000          0         0        11     879454.4                1              6             1              44

Take CustomerNo=1 as example,M1T1=2 and M1A1=500.4 means this customer had  number of transaction 2 and amount of transaction 500.4 at merchant type M1,
There are total 20 merchant types and 490000 Customers.

Now need to do Customer Classification and identify Consumption Habit.

Which ways will you choose? And can you share the code in SAS?

[Reeza]

Do you have SAS EM? This sounds like Market Basket Analysis to me.

SAS Base doesn't have the mechanism to do this well.

[summy]

Yes, I have SAS EM and SAS/STAT

[Reeza]

Then I'd probably recommend walking through a tutorial on Market Basket Analysis either in the SAS EM documentation and then move on to your data.

Also consider reading up on what MBA is, if you're unfamiliar with the concept.

[summy]

Walked through tutorial months ago,only found the result SAS giving like Cluster is a little specious.
What do you think?

[Reeza]

MBA isn't cluster analysis AFAIK and does require SAS EM. This whole field of customer classification is a bit new and still in flux in my opinion. I haven't tried any ML algothims but I know a lot of people do.

[summy]

I meant proc cluster; in SAS/STAT. Anyway,let's find out a way, an executable way.