The kind of raw data we are talking about might be data from a care home covering, say, a three month time period.
The challenge comes when you realise that for one resident in that time period there could be 50+ data streams
and 7,000+ rows of data.
Not only that, but the data would contain a mix of care activities carried out
and care observations too.
Each data stream is converted from its raw form to a processed form based on time windows that make sense for the data stream.
This graph shows fluid intake based on three-hour time windows.
The graph shows the general pattern of fluid intake as between
200ml and 400ml per three-hour window. That's from 06:00 to 18:00 and with occasional fluid intake later in the evening.
After basic processing, data from each data stream is passed to an AI-based pattern learner which learns
the normal pattern for each data stream.
The graph shows the normal pattern for fluid intake. It shows that the resident normally consumes 400ml of fluid in
each three-hour block from 06:00 to 18:00, but that it is not unusual for this to be 300ml or 200ml.
Fluid intake outside these times is less common and also in less quantity.
This chart shows the individual indicator for fluid consumption.
Most days past the 30-day learning phase are in green, indicating that they are similar to the learned pattern.
There is a gentle trend upwards in score as well as an increasing number of amber and red scores over time, which indicates
a potentially emerging problem.
Find out more...
We believe that AI and machine learning are the keys to turning big data and deep data into
personalised, predictive insight to empower preventative approaches to health and care. We’d love to
talk to you and show you what we could do with your data.