The inaugural Predict 2015 conference brought key players in business, government and research to Dublin to share learnings and insights on data modelling and predictive analytics. Here’s what we learned from them.
More than 350 professionals in data and analytics attended Predict 2015 in Dublin’s RDS, and plans are already underway to continue the event with Predict 2016.
Event sponsor the Aon Centre for Innovation and Analytics plans to host a live revision webinar of the conference in its Dublin office on 21 October and an e-book providing an overview of the event will be provided following that.
In the meantime, here’s what we learned from some of the event’s headline speakers.
1. Data on its own isn’t enough
The founder and CEO of Creme Global, the host of Predict 2015, presented his keynote during the ‘Lessons from Data’ session. McNamara compares the current explosion of data with a Wild West land grab, but makes the distinction that – unlike land – data is limitless.
But, McNamara clarifies, “Data, on its own, isn’t enough.” Data needs analysis, interpretation and presentation and, while tools to achieve this are widely available, they are often inaccessible to those without data-scientist-level training.
These resources can present a barrier-to-entry for small businesses seeking to gain insights from data, but Creme Global is working on a platform to help democratise the process.
2. Packaging is important
As the new CEO of the Aon Centre for Innovation and Analytics (ACIA), Todd Curry oversees a staff of more than 100 deilvering data-driven insights to clients who sometimes demand perfection.
In response to those requests, Curry asks, “Why won’t an imperfect answer do?”
The key to a client’s understanding of the ACIA’s work, however, is communication, and that’s why packaging insights learned from data in an understandable format is mission-critical.
3. Whether it’s your models or your staff, teamwork is the secret weapon
John Elder, president and founder of Elder Research, recognised at Predict 2015 that data science is now having its moment in the spotlight, but he’s not a believer in the unicorn-like myth of a multi-skilled data scientist working solo to deliver results.
For Elder Research, teamwork is key to bringing the necessary skills and differing perspectives to data analysis. This same concept can be applied to the selection of data modelling techniques, as Elder advocates for a ‘board of directors of models’.
This “secret weapon” of combining different modelling methods can help determine the best approach in an effective way.
4. Avoid overconfidence bias
Dr Constantin Gurdgiev, adjunct assistant professor of finance at Trinity College Dublin, has witnessed dramatic improvements in financial reporting over the past 15 years, but he sees a significant challenge brought on by the data explosion: overconfidence bias.
Because of the sophistication of tools that capture and analyse data, we believe the insights to be scientific – but Dr Gurdgiev says that we should imagine that term in quotation marks.
Analysts need to understand their limits and apply a certain craft to their interpretation of data, advises Dr Gurdgiev. As he puts it, “Data is not the end of it all. Data is just one of the tools.”
5. Data science is not for data scientists alone
Brian O’Mullane believes that advances in data science and the tools used won’t just be transformative for other industries, but for data science itself.
“By making better tools available to everybody, we can potentially disrupt data science itself and let other people beyond data scientists work with this data, and build beautiful web apps that they can potentially allow decision-makers to quickly analyse their data on,” he said.
For O’Mullane, this democratisation of data science makes perfect sense when you look at how software development has evolved from a specialist subject for computer science graduates to something that people with varied skillsets can now collaborate on.
6. There is no artificial intelligence without data
As well as assuring us that artificially intelligent robots are (so far) only being built to assist and not compete with (or destroy) humans, IBM’s Duncan Anderson explained how data is critical to AI.
“Without data it’s very hard to be intelligent,” said the CTO of Watson Europe.
“If you have all sorts of systems where you can answer questions, where you can interact with things – without the data, you are just having a nice chat. The data makes it personal, makes it relevant, makes it specific. [It] gives you answers to questions.”
7. Just as there’s good data and bad data, there are good biases and bad biases
There are two sides to everything in data, according to Dr Mike Bolger, former toxicologist with the US Department of Health & Human Services. You can have good data and bad data, depending on the quality. And even good data can suffer from bad interpretation.
Interpretation of data is often skewed by bias, and even in this Dr Bolger sees duality.
“There are good biases and there are bad biases,” he said.
“People who feel strongly and passionate about anything – well, that’s a bias. It’s not necessarily a bad thing, but it can have its downside if the bias is used in a way where you knowingly misrepresent a set of data. That, to me, is what I would call bad bias.”
Siliconrepublic.com’s Data Science Week brings you special coverage of this rapidly growing field from 28 September to 2 October 2015. Don’t miss an entry worth your analysis by subscribing to our news alerts or following @siliconrepublic and the hashtag #DataScienceWeek on Twitter.