|
However, within a few weeks of working with 11Ants Analytics and reviewing preliminary results we could see that their approach actually yielded encouraging results within a short period of time.'
Solution
Fisher & Paykel Healthcare engaged 11Ants
Analytics and issued them the challenge of
finding patterns in supplied data.
11Ants Analytics mobilized its team of leading
data scientists. The combination of 11Ants
Analytics proprietary technologies and knowhow
did indeed discover new patterns in the
data. This was no regular simple statistical
analysis exercise, rather careful deployment of a
combination of complex machine learning
techniques, complex data transformations and
highly advanced mathematics. 11Ants Analytics
worked closely with Fisher & Paykel Healthcare’s
engineers over a period of several months.
The engagement followed 11Ants Analytics
standard four step process of engagement for
professional services:
- Plausibility
- Experimentation
- Model
Development
- Deployment
This methodology serves as an effective means
for clients to exploit advanced predictive
analytics while significantly de-risking the
process. A brief description of the process
follows:
- Plausibility - plausibility requires
establishing if there is a plausible correlation
between the data available and the thing
required to be predicted. This does not
require any sophisticated analysis, rather is a
matter of consultation between 11Ants
Analytics and Fisher and Paykel Healthcare. In
this case it was established that it was quite
plausible (though as yet unproven) that there was a relationship between the data and
respiratory states.
- Experimentation - this involved doing
some preliminary analysis of the data to
establish what the likelihood of success was.
It was determined that useful patterns did
exist in the data, and that it was worth
continuing on towards full predictive model
development.
- Model Development - this component
involved more diligent analysis of the data with
view to build the best predictive model(s)
possible. Often this may involve gathering
additional data, different data transformations,
and increased algorithm evaluation. This work
is now significantly de-risked based upon the
results of the experimentation phase. 11Ants
Model Builder and other proprietary
technologies featured heavily in both the
experimentation and model development
phases.
- Deployment - this is where the predictive
model gets ‘operationalized’. In this case it
was determined that the most logical solution
was to code the resultant algorithm out into
C# so that this could be flashed onto a chip
that would be embedded in the device.
Benefits
At the completion of the engagement Fisher
& Paykel Healthcare were provided with a PC
executable algorithm that was able to predict
the nature of the observed respiratory
phenomena.
The end result is a new competitive
advantage to Fisher & Paykel Healthcare in a
several hundred million dollar market and a
very satisfied customer.
The final word goes to
David Whiting:
‘The team at 11Ants Analytics has greatly
assisted my team’s R & D. Their innovative
techniques and software allowed large
amounts of data to be quickly analysed and
provided rapid proof of concept. The result
will be a substantial improvement to one of
our products.’
|
