Data Analytics Newsletter #11
October 2019

The Data Analytics Practice Committee (DAPC), the Young Data Analytics Working Group (YDAWG) and the 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.

The month has passed quickly! In this edition, we return with two must-attend Data Science events hosted by the Actuaries Institute, the questions you have to ask MI vendors, and a freebie from Microsoft!

Contents

1. AI and Data Science Masterclass
2. Journeys to Data Science
3. Artificially Misleading
4. Dodgy AI YouTuber exposed
5. Three questions you must pose to machine learning vendors
6. Free visualisation tool from Microsoft 
7. Winning Solutions
8. Academic Corner - Gated Recurrent Units
9. Editor's Note

AI and Data Science Masterclass

For those interested, the Actuaries Institute is offering an intimate one-day MasterClass in partnership with AlphaZetta aimed at current or aspiring leaders of AI / machine learning teams and functions. Presented by Dr Eugene Dubossarsky - who founded Data Science Sydney (DSS) with its 6,499+ members - and has trained hundreds of aspiring data scientists in the corporate, startup, government and health sectors. Find out more and register today for a seat on Thursday 28 November in Sydney (places limited!).

Hong Kong Insights session - Journeys to Data Science

A panel of five Hong Kong based actuaries will share their personal journeys into the realm of data science, followed by a Q&A session. Find out more about the session and register today for this event on Wednesday evening 13 November, at EY Hong Kong!

Artificially Misleading

With data analytics being so popular right now, less reliable operators are cashing in, and it’s worthwhile to include in this newsletter a reminder for a healthy degree of actuarial scepticism and vigilance. A neural network with 300 neurons can be a great model, but sometimes one neuron is more informative. 

Dodgy AI YouTuber exposed

A popular YouTube AI educator has been caught copying materials without citation and setting misleading expectations for a paid 10 week online course on data analytics. Shortly after, he was exposed for plagiarising a quantum computing research paper, where he bizarrely replaced the term “quantum gate” with “quantum door”.

Three questions you must pose to machine learning vendors

Dr Celeste Fralick, Senior Principal Engineer and Chief Data Scientist, shares her top three questions technology consumers must ask, as well as what makes a good and bad answer. Check them out here.

In addition, here are the questions I'd also ask:

1. What benefits does this ML approach have over the ‘one neuron’ approach?
2. When it gets it wrong because no system is perfect, how does it avoid bringing my  whole organisation into disrepute?
3. How much does it cost?

Free visualisation tool from Microsoft

Microsoft have released for free and open-sourced their SandDance visual data exploration tool.

It can be used from your browser (on desktop - not so great on your mobile screen for those reading this on the train) or as a plugin within:

• PowerBI
• Azure Data Studio, a free SQL query tool for MS SQL Server and Postgresql
• Or Visual Studio Code, one of the best editors for coding in Python.

They all look great. Check them out!

Winning Solutions on Kaggle

I started my data analytics journey by reading winning solutions of Kaggle competitions. A competition on fraud detection just ended and there is some interesting content. The results will surprise you! 

• Notebooks - with exploratory data analysis
• Top models

Academic Corner - Gated Recurrent Units

The gated recurrent units is a type of neural network used to model sequence data (such as natural language or time series).

The underlying concept of a gate is fairly intuitive. You have a multiplier that goes from 0 to 1. If it is zero, nothing goes through because it gets multiplied by zero. If it is one, the signal goes through in its entirety. In the context of a GRU these are switches to decide whether to “remember” (update) or “forget” (reset).

The maths can be a little intimidating, but you can find an article that explains it with diagram here.

Disclaimer

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