30 November 2021

Data Scientists Make Their Predictions for 2022

2022 is fast approaching and the data science industry is showing no signs of slowing down, despite a year of rapid changing across the way people work, their compensation, benefits and a continually heating market that will face skills shortages well into 2022.


With this in mind, we asked four data scientists from the BBC, Audigent, Abrdn and the FinTech sector what companies and data scientists should expect from 2022.

These are their predictions:

Christina Boididou, Data Scientist, BBC

 

In the recent years, more and more applications around healthcare are popping up. Data Science can play a key role to solving problems in the health domain as collective use of data can shed light to unknown areas and be an assistive tool to health professionals. Especially now, with the pandemic situation, data usage was proven to be very important to understand how the virus is transmitted, what are the symptoms in different populations, etc. I would expect industry in the near future to focus more and more on solving problems around health.

 

Davide Anastasia, Head of Data at Audigent

 

Hard to say: it’s a landscape in constant motion and very hard to predict.

I am expecting new AI/ML frameworks to be developed, but I think those will be mostly evolution of what we have now, instead of revolutions. There will also certainly be improvements in tooling that will help us standardise the way we develop new models and push them into production: the lack of a “canonical way” seems to be an important hurdle in the road of completion of data science projects. Finally, I am expecting cloud vendors to keep improving on their data offering (DW, ML training and inference, data pipelines and so on), as the level of maturity is slightly different across the major vendors.

Gavin Allan, Data Scientist, FinTech

 

A key challenge in the industry has always been the need for labelled data – which is often time consuming and expensive to obtain. I expect the capabilities of semi-supervised learning – which utilises labelled and unlabelled data – to develop significantly over the next year as we look to build high performance models without the need for costly data labelling exercises.

As models have become more complex, explainability has inevitably suffered but with increased interest from industry regulators, as I talked about above, I expect developments in this area to continue to grow significantly.

Martin Thorn, Head of Data Science at Abrdn

 

I think 2022 is when more and more big corporations start asking what they are getting for their ML investment. Too many companies are spending big and getting very little in terms of commercial output. Granted, I've been saying that for a few years now but it’s going to happen soon. Almost certainly.