Data Smoothing

anilgvdbm — Fri, 01 Feb 2019 06:53:47 GMT

Hi Experts,

I have one question regarding the data smoothing in SAS. let's say i have 4 years data for 10 million customers.

Sample data looks like this

Cutomer_ID	Year	values
2	2014	40
2	2015	24
2	2016	14
2	2017	26
4	2014	1
4	2015	18
4	2016	9
4	2017	8
5	2014	30
5	2015	20
5	2016	10
5	2017	12

So from the above table, i have to create risk tag for each customer based on their four years score. but in this case, if the customer gets a good score in years and a bad score in 2 years.

So I need to classify based on the smoothing technique in SAS.

Please, can i get any code reference for this?

Thanks in advance.

Regards,

Anil

Re: Data Smoothing

Rick_SAS — Thu, 07 Feb 2019 18:32:33 GMT

Four years is not very much data for smoothing. I think the best technique would be a simple one such as an average or a weighted average. If you use a weighted average, you probably want to assign more weight to the most recent data, such as

w = {0.125, 0.25, 0.5, 1};

or standardize by defining

w = w / sum(w);

If you use this to develop a score for Customer_ID = 2, the score for that customer would be

Score = 0.125*40 + 0.25*24 + 0.5*14 + 1*26 = 23.47

For a reference, you can read about various kinds of moving averages and how to compute a moving average in SAS. But if you know that you have exactly four observations for each ID, you can use BY-groups processing and the FIRST.ID and LAST.ID variables to add the value to the data set.

topic Data Smoothing in SAS Data Science

Data Smoothing

Re: Data Smoothing