topic Pattern Recognition in QlikView in QlikView

Pattern Recognition in QlikView

alexis — Wed, 10 Dec 2014 15:40:08 GMT

I have an application that reports on credit card transactions using credit card "terminals" - these are the handheld machines that you type in your PIN when you pay using a credit card. Each of these terminals has an ID so it is uniquely recognised.

The requirement is to identify terminals that are no longer used or are out of service. Identifying this fact is a little tricky as some terminals are used every day (so it's easy to identify when they are no longer used) but some are used during the week only and some weekends only.

Are there any tools or does anyone have any ideas regarding recognising the pattern of use (from the transactions) and identifying a "break" in that pattern?

Any assistance would be greatly appreciated.

Thanks in advance

Alexis

Re: Pattern Recognition in QlikView

MarcoWedel — Wed, 10 Dec 2014 23:10:44 GMT

If(Today()-Max(Date)> x, 'unused', 'used') as terminalusage

with x being the maximum allowed number of days since the last usage (date).

Re: Pattern Recognition in QlikView

MarcoWedel — Wed, 10 Dec 2014 23:17:37 GMT

you also could create a usage "pattern" per week like

Sum(Pow(2, WeekDay(Date))) for all Dates having transactions using one terminal.

Rank 1 of this pattern count over the last x weeks could be the usual pattern of a specific device and non matching patterns could be marked as a change.

hope this helps

regards

Marco

Re: Pattern Recognition in QlikView

alexis — Wed, 10 Dec 2014 23:50:55 GMT

Thanks Marco - you are ion the right track as each terminal has a different length of "inactivity".

Am I understanding correctly that,for example if a terminal works weekend only then,for a specific week:

2power5 +2power6 = 32 +64 = 96

What you are suggesting is to calculate this on a week by week basis and see how closely it is to the first week by ranking? Can you give an example - sounds really interesting..

regards

Alexs