People usually start with a mix of simple thresholding(amounts over the usual, eg) and rules-based classifiers(transactions in more than one geographic area within a small time frame, eg), then work on training a classifier that factors in more and more variables as appropriate. See the section on the website about classifiers for more about the available methods.
There is also a Jira ticket about incorporating some of Ted's work which if you're interested you could work on. Best Andrew > On Jan 3, 2015, at 7:33 PM, "Zhangxiaopeng (Eric)" > <[email protected]> wrote: > > Hello, > > I am a newbie to big data analysis. Recently, I receive a request about > preventing fraud with big data analytics. > > I want to know which models and algorithms can be applied in the preventing > fraud field ? And can I use mahout to support these modules and algorithms ? > > I am appreciate if someone give some tips about it. > > > BR > > Eric > > > >
