As the first step in the decommissioning of sasCommunity.org the site has been converted to read-only mode.


Here are some tips for How to share your SAS knowledge with your professional network.


Data mining models for improving card fraud detection

From sasCommunity
Jump to: navigation, search

Author: Abraham Nieto

Card fraud is a headache for the Banks everyday, and the solutions offered in the market ensure that the support of the tools is based on analytical procedures or analytics. All companies on the fraud detection solutions market offer us a solution supported by a model like neural network and they talk about a score created by themselves which it will give us a great assertiveness, but they never tell us how the black box or model is built(What conditions?, what variables were used for calibrating and creating?) instead of that they teach us how to use the tool for creating rules (if... else structures) and sometimes they don't know that the users have a datamining background and they know how to create datamining models of couse! and further they don't know that them own tool could be used for writing mathematical expressions therefore datamining models too. except sas fraud solutions, the rest of them only show us how to write rules and none show us how to write an expression that could be associated to a model expression. in a bank it was discovered how to write a neural network instead of a rule and for improving the score that the provider sells. finally, it's important to demand a lot of explanations related to customer needs.