Raiz Invest partners with UNSW to advance machine learning
Raiz Invest has partnered with UNSW and with support from the NSW Government to develop a machine-learning recommendation engine for its Raiz Rewards.
The new technology will supposedly recommend to Raiz customers more relevant offers and brands.
It’s designed to provide a more seamless, personalised experience for customers when navigating more than 165 brands on Raiz Rewards.
With a customer base of more than 179,000, skewing towards the under-35 demographic, the experience provided by machine-learning will supposedly better address the demands of this audience.
Greater personalisation and relevance
The machine-learning engine powering the Raiz Rewards platform was designed and built in collaboration with UNSW’s Lina Yao, a leading machine-learning researcher with extensive experience in recommender systems.
Raiz was awarded a matching $15,000 TechVouchers research grant through the NSW Government’s Boosting Business Innovation Program to work with UNSW on the project.
Minister for small business John Barilaro said through the ‘Boost’ program the NSW Government has provided $18 million to the state’s 11 universities and the CSIRO to work with businesses to create great new ideas and products.
“This is about entrepreneurs tapping into the top-notch research provided by our university sector,” Barilaro said.
“As a government, we want to do everything we can to create the right environment for people to have the confidence to launch a startup and then get the support they need to succeed.”
“As a result, NSW is now home to nearly 49 per cent of Australia’s startups.”
What has been developed by Raiz and UNSW combines two common approaches, used by companies such as Amazon or Netflix, to a recommendation engine to give a higher level of personalisation and relevance.
This was achieved by constructing two parallel neural networks, where the predictions from each were weighted and summed up, to provide a final prediction.
The resulting model should give more relevant recommendations, without needing to incorporate complex information about either the user or the product.