Data Analytics Newsletter #18
May 2020

The Data Analytics Practice Committee (DAPC), the Young Data Analytics Working Group (YDAWG) and the Actuaries Institute are pleased to bring the latest in the world of Data Analytics to your inbox, and to share some of our recent work with you.

In this issue, we revisit COVID-19 related issues in data analytics, and highlight learnings which may have applicability beyond the pandemic.

Contents

1. The challenges posed to data science amid COVID-19
2. Extrapolations and Cubic Curve-Fitting
3. Contact Tracing and Privacy
4. Rob Daly's path to machine learning

The challenges faced by data science amid COVID-19

COVID-19 has understandably drawn significant attention from the data science community. As a matter of public interest, data science can help inform discussions about public safety, and communicate important messages such as '#FlattenTheCurve'. At the same time, data science issues of data quality, inappropriate modelling approaches, or ineffective visualisations can lead to poor predictions and mis-informed decisions.

This piece in the Elsevier Public Health Emergency Collection provides some highlights in the modelling, visualisation and data sharing that has taken place as well as suggestions to prevent misinformation that may be relevant for both data scientists and actuaries looking at the issue.

Read more.

Extrapolations and cubic curve-fitting

Unfortunately the rush to deliver helpful insights on COVID-19 can easily lead to poor models. One such model on US COVID-19 mortality published on Twitter is critiqued by the Statistical Modelling blog of the University of Columbia in this short post.

It links to R code replicating the projection, and highlights how using a cubic projection may not be an accurate representation of the underlying process, leading to overfitting and a sharp drop towards zero – until it suddenly goes parabolic (technically cubic).

The same blog also provides a critique on another poor visualisation of COVID-19 development with lessons in effective graphical communication.

Read more.

Contact tracing and privacy

Contact tracing to identify those at high risk of COVID-19 due to having come in contact with an infected individual has emerged as an important strategy in containing the spread of the virus. However, it is also important to address privacy concerns in a post-Snowden world.

To promote transparency, Singapore released its contact tracing app as open source, and the  Australian Government’s 'COVIDSafe' draws on that model. It is encouraging how the global community is sharing knowledge on how best to combat the disease.

However, privacy advocates such as Mozilla have described how a decentralised system may provide mathematical guarantees of privacy that are stronger than the legal protections that have been offered.

The discussion is another reminder that just as with any other data collection, the community expects data to be used for only the purpose it is collected.

However, let not perfect be the enemy of the good: over five million people have downloaded the app, and as our President Hoa Bui says – it is the neighbourly thing to do.

For further discussion of data privacy issues in a more general context, a recent podcast with DAPC Convenor Bartosz Piwcewicz and committee member Amanda Aitken discusses DAPC’s role in navigating issues in the ethical use of data.

Read more.

Rob Daly's path to machine learning

Rob Daly has had a long and successful career as a Chief Risk Officer, a Chief Actuary, a reinsurer and a consultant – but one day he became intrigued about some “mysterious” “non-parametric techniques”. So, he did a course on Machine Learning, learnt how to tell the difference between a shirt and a dress, and now he is the Institute’s Data Analytics Strategy Consultant!

In this article he shares his pathway to understanding how neural networks and other data science techniques work and describes how actuaries have a great foundation to learn the coding, statistics and machine learning skills to become successful data scientists.
Read more.

Disclaimer: The Institute wishes it to be understood that any opinions put forward in this publication are not necessarily those of the Institute.