Skip main navigation
We use cookies to give you a better experience, if that’s ok you can close this message and carry on browsing. For more info read our cookies policy.
We use cookies to give you a better experience. Carry on browsing if you're happy with this, or read our cookies policy for more information.

Skip to 0 minutes and 5 secondsSo all the technologies, we've discussed, so far, transport that huge amount of IoT data-- our things, in general, generate. Data analytics programmes, however, is what gives that data some life. It is this knowledge and wisdom we can extract from I.T. data, which makes the Internet of Things so interesting. Data analytics algorithms vary significantly in capabilities and scope. So there are algorithms which find us the 'knowns knowns.' Then there are algorithms which are able to extract the 'known unknowns.' And as of very, very recently, there are algorithms which are even able to extract the 'unknown unknowns' from our data sets. An example of the first class of algorithms, the 'known knowns,' is simple linear regression.

Skip to 0 minutes and 52 secondsSo for example, assume we instrumented a shop in the high street with sensors which are able to measure the amount of goods on the shelves. If we now plotted that amount against time, linear regression would allow us to establish the best fit for how quickly the shelves are being emptied. An example of the second class of algorithms, the 'known unknowns' is machine learning. Now here, sophisticated algorithms are able to arrange data into clusters of prior specified characteristics. In addition, these algorithms are able to predict most likely events to occur. With our previous example, machine learning would be able to establish which products are being taken off, at what rate, and therefore enable better supply chain decisions.

Skip to 1 minute and 41 secondsNow the third class of algorithims, the 'unknown unknowns,' is currently pushing the boundary of what is possible and artificial intelligence. These deep-learning algorithms some of which also rely on machine learning are able to get insights from data which we humans even didn't imagine existed. Coming back to our example of a smart shop, deep-learning analytics may reveal why certain products fly and others don't. So what's the aim of these analytics approaches? Why do we do this? Well, generally we’re after two ends of the insight spectrum, really. On the one end, IoT and data analytics allow us to detect anomalies in real time that let us trigger an alert of something going wrong or about to happen.

Skip to 2 minutes and 29 secondsFor instance, using our example if we detected that a specific product ran out an alert could be sent to the shop owner to order more stock. That alert could even be sent to the supply chain without the shop owner even noticing. Now on the other hand, collecting all the data and crunching it over time allows us to detect long-term trends, which in turn allows us to construct certain policies. For example, imagine we observe that a specific product does not sell in winter but is really-- really popular in the summer. The shelf stocking policy could then be adapted to these long-term patterns. Where and how to implement IoT data analytics algorithms is something which depends on the context.

Skip to 3 minutes and 12 secondsSome very difficult trade-offs typically need to be struck, since our IoT devices are highly constrained on resources. Should I use a lot of computing energy and memory in my IoT device to crunch all that data and then only send the final result to the IoT data service? Or should I use a lot of energy to submit all the raw data and rather have the calculus done in the back end server? Now that trade-off is heavily dependent on the IoT application at hand, and generally is not an easy call to make. Now to conclude. With a rich set of data analytics algorithms, the IoT is generally a great contender of big data approaches.

Skip to 3 minutes and 56 secondsIt is indeed the IoT which gives us sufficient temporal and spatial data granularity to obtain meaningful real time and long-term insights to make various industries and processes more efficient and more effective.

Data analytics

In this video Mischa introduces the capabilities of data analytics in the Internet of Things, and discusses some examples of different analytics approaches.

Share this video:

This video is from the free online course:

The Internet of Things

King's College London

Contact FutureLearn for Support