NBN Co will turn to machine learning and big data to identify patterns in the way faults occur in the installation and activation of new connections.
The company is set to create a virtual ‘tech lab’ with personnel in Sydney and Melbourne.
It will use a range of technologies and algorithms – all yet to be decided – to solve problems experienced by end users as new connections to the NBN are made.
In particular, NBN Co is hoping to use algorithms to work out whether it needs to physically attend a premises to remediate a fault, or whether the job can be performed remotely.
The size of the data set that NBN Co hopes to use to train the algorithm wasn’t immediately clear – the company said it had access to “survey” data from consenting users but a spokesperson was unable to provide further details.
Likewise, how NBN Co plans to achieve results through the lab is equally opaque.
The network builder is interested in a laundry-list of technologies, spanning Apache technologies like Spark, Kafka, Flume and Cassandra, as well as open source deep learning platform H2O.ai.
It was unable to confirm what it already had in its IT or data environments versus what it would have to test or deploy for the first time.
NBN Co is increasingly looking for ways to improve the experience of end users connecting to its network, after a rough period through the middle of 2017 where it faced sustained pressure to improve the customer experience.
The company this week said it would work with customers of its fibre-to-the-node (FTTN) network to improve their in-home wiring, and hopefully increase download speeds.
Technology to be introduced next year aims to further stabilise FTTN connections.
The company has also set up a centre for information quality and effectiveness (CIQE) so that all parties involved in a connection are on the same page when talking to an end user.
And it has been involved in a months-long process with its top retail service providers to iron out problems, as well as offer them deeper discounts to encourage the purchase of more backhaul.