What Does The Rise Of Machine Learning Mean For Asset Managers?
Machine Learning (ML) is enjoying rapid uptake across the asset management sector, with many suggesting that it could reverse the trend towards passive investment funds.
Can it? Will it? Is it even a good thing? As with any new technology, the truth is that it won’t work for everyone – and there are both good and bad things about ML.
But let’s start with the good:
The beauty of ML is its ability to automatically analyse huge amounts of data to reveal useful information and make predictions based on defined targets. And, instead of having to precisely follow instructions, ML is able to self-adjust to improve accuracy.
There are three main benefits of using ML:
It can mitigate against the negative effect of human bias when it comes to investment decisions. The fact is that humans are often irrational, which effects decisions – for example by using confirmation bias (when new information is looked at simply to confirm existing beliefs) to justify actions that turn out to be wrong. ML can be used to analyse investment decisions and detect if any bias is involved, which means the issue can then be addressed with the people in question.
It has the potential to outperform equities by finding new patterns in data sets. For example, by analysing historical decisions to ascertain whether they led to to good or bad stock performance, which can obviously be useful in guiding future decisions.
It introduces the ability to analyse new forms of data – e.g. images or sounds. This has potential to inform investment decisions, for example by analysing satellite images of agricultural land to predict crop yields.
So far, so good. But ML can also have a few significant flaws:
It makes predictions by analysing data derived from past events, so it won’t predict events that aren’t linked to what has happened before.
The patterns it uncovers can’t necessarily be used to make profitable investment decisions.
Deficiencies within the algorithm, or the data sources used to train the system, may result in ML displaying its own biases.
And asset managers must be very aware of these problems and try and mitigate against them. This can be achieved partly by employing experts in the field – data scientists – to set up and operate your ML system, so that it can be properly instructed in what to do and you have someone in your team with the necessary technical experience to realise when things need changing.
Also, you must ensure that you run computer programmes that are sophisticated enough to work with ML. You can’t continue to rely on Excel spreadsheets.
In short, you will come unstuck if you expect ML to do everything for you. Although it can vastly improve the quality of data analysis, it does not replace human judgment. So, you still need capable people in your team to install, manage and guide ML, and critically evaluate its output.
