Transcriptions
Note: this content has been automatically generated.
00:00:04
who's not able to be here today and i'll try to do my best to
00:00:07
present his project um and we talked about
00:00:11
a lot you've talked about uh today about
00:00:15
how to creating classes difficult and this is kind of the high level case study
00:00:21
of how it's actually creating custom transparency as act absolutely necessary
00:00:27
for the success of the projects uh uh especially this one
00:00:33
so all's started by presenting the speaker who's not here
00:00:38
gary see the has a double masters economics and management
00:00:44
and computer science and you didn't master here yep
00:00:48
and you was uh students well what's doing is masters
00:00:52
with a circle mature and and i'm a machine learning a id that could lead to an
00:00:57
never that background from it and and and that that's interesting and and and that s. b.
00:01:03
uh so quickly the context and told me to have four of those of you who maybe don't know us this well
00:01:09
uh where are um ensure us wasn't sure
00:01:14
uh and we ensure people globally so of course healthcare probably know that
00:01:20
but also twelve and a is small companies
00:01:24
medium companies larger companies and self hatred printers
00:01:29
the few numbers so we have about one point three million
00:01:33
uh customers uh and it's twenty eight million a company clients and
00:01:41
as you can imagine it it's a big operation so if we average
00:01:46
over here we have about three thousand your clients coming in every year
00:01:50
we get a round six thousands of the calls ten thousand emails let's
00:01:56
uh what's important for this talk is seventy seven million uh thousands
00:02:01
uh invoices that come in every day uh and we reimburse from all
00:02:06
these invoices about two point through the twenty three point five million swiss francs
00:02:11
so as you can imagine with that kind of of volume of process
00:02:16
it's very difficult it's a two identified a
00:02:20
fraud and identify the cases where some people
00:02:22
are trying to gain the system to gain us and to get money that isn't do um
00:02:31
the the context before we started this project was uh that
00:02:36
we knew kind of from the literature from reading articles that
00:02:40
in in general the company like us five to ten percent of claim are not necessarily fraud
00:02:46
but could be suspicious and would raise an
00:02:49
alarm and we have a a menu deterministic rules
00:02:54
uh based on the legal framework based on our
00:02:57
contracts to catch most of these cases but then of
00:03:01
course there are some cases that pass these the deterministic rules which we don't really know how much they are