Machine learning still requires human element
Machine learning has become a fundamental part of active management, but despite the practical implementation it has, human error-checking was still required.
Jean-Christophe de Beaulieu, head of investments at Acadian Asset Management, said for one piece of information that finds its way into a model you have 10 that are disregarded.
An example of this not working was a joint venture with Microsoft to use the search engine Bing.
“We were trying to search through the words and when they were connected to some companies, whether that was conducive to performance or underperformance,” de Beaulieu said.
“If you think about the brain power of Microsoft and the brain power of Acadian combined, we looked at the data sets for six months, we tried to sweep through the algorithm and extract some value and we didn’t come up with anything.”
They had introduced more machine learning as a way of assessing the typical momentum and price movement of a particular stock.
“The machine learning algorithm can look into these variations and pick the ones which are robust across a number of regions and we can make the best use of it,” de Beaulieu said.
“Machine learning is something we’ve been using for a long time, but now there is more exposure in different parts of our model as far as expressing some of the factors."
Recommended for you
Legal profession-focused super fund legalsuper has partnered with Link Group to launch a cloud-based digital platform.
The number of investments being made online has fallen significantly amid the market downturn and the threat of rising inflation and interest rates.
Fintechs will now have a regulatory sandbox that will allow them to test new products and services for 24 months without obtaining a financial services or credit licence.
Publicly-listed platform provider Netwealth has continued to assert its leadership – this time in terms of digital wealth applications.
