Data Intelligence, Business Analytics
Hi Every one,
has anyone done support vector machines using SAS? I am currently building classification models for fraud detection and prevention. would like some advice on what approach i should use to implement support vector machines.
anyone done some work on statistical techniques for fraud detection and prevention.
Tags:
Permalink Reply by Tomas Keller (formerly Ohlson) on March 3, 2011 at 9:42am Hi,
when I used SVM in SAS there were only two different kernel functions, is it still?
If I remember correctly I transformed all variable values into the interval [0,1].
I worked in a project where we didn't have any confirmed cases of fraud but wanted to monitor transactions that could be fraudulent. I successfully used peer-group clustering to observe behavioural changes among the accounts.
Regards
Tomas Keller
Permalink Reply by Mark on March 3, 2011 at 10:58pm thanks Tomas,
there is no support vector machine procedure in SAS at the moment. Many SAS users believe it should be released sometime this year.
Tomas, I have just started doing research on the classification methods,the bank i work has never implemented these methods before and my plan is try and cover peer group analysis at some point in time. at the moment at working on the regression approach, then will look to work with non parametric methods.
do you still remember how you went about peer group analysis? i have read an atical by David Hand and his research team. let me have your approach on how you went about doing peer group analysis. thanks in advance.
Permalink Reply by Tomas Keller (formerly Ohlson) on March 4, 2011 at 1:18am Hi Mark,
SVM should be accssible in SAS EM not sure if you can find it in Base SAS.
I used the ideas from the article Unsupervised profiling methods for fraud detection
by Bolton and Hand and derived my own method to compare the difference between the peer-group and the target.
/Tomas
© 2013 AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC