Predictive Analytics World — totally excited

Looking for data science conferences you can’t get by the Predictive Analytics World. More than 650 data enthusiasts participate the Data Driven Business Conference, including the Predictive Analytics World. The 3 conferences focus on Online Marketing and Analytics.

Over the last 8 years in different companies the discussion about conferences ended with the same question: WHY? NO! But this year at the Digital Unit I asked again and emphasise how important conferences are, including networking, be up to date and open your horizon. This time the answer from my boss were YES! but let’s discuss which conference suits to my needs and bring also a value for my work.

Berlin, Berlin, we’ll travel to Berlin!

My colleague and me traveled to Berlin, excited about the event and curious about the sessions. At breakfast you could see a lot of people with the DDB Badges, not these young Data Science Genius in T-shirts, no hipster, traditional people of every age. On the Expo we also couldn’t find a t-shirt booth, different to the Tech Conferences of my colleagues. Let’s focus on the talks!

Predictive Analytics World

The conference offers a wide range of topics, great performance or rather sober, mostly high level, mainly in german. The audience contains online marketing experts, high level management, analysts and data scientists, divided into talks for the whole audience and special talks at the same time.

The first talk concerned with the topic “The centrality of a detailed understanding of your audience”. Haile Owusu from Mashable gave a overview of the value of content. With deep learning they created an impact score to predict the advantage of content.

7LYTIX talked in the next session about changes in the Retail by using deep learning and predictive analytics. They are using several methods (deep learning, regression, time series decomposition) to predict the sales and using image classification.

Both talks were high level and more an introduction of their topic.

My former colleague showed me sketch notes as a useful tool to filter the important facts and better focus. I love to use it for the conference. The first special talk treated with personalisation of display advertisement at OTTO.

“Machine learning in industrial data spaces” or there are unknown unknown. The industrial data space association contains standards for one ecosystem and one infrastructure for all companies involved. The presentation focused on guidelines to to prepare companies for machine learning.

The talk about “how predictive modellers should use data to tell data stories” was very interesting. We all know bedtime stories but how can we visualise data to tell stories. We don’t have time to create masterpieces. The visualisation depends on the info, but we should “save the pies for the dessert” and create scatterplots for the analysts not for the bosses. At the end you should keep it simple, be transparent and quickly.

Of course there are provider of tools, software and consulting. Hitachi for example showed the Pentaho ML.

My highlight of the first day was the talk about “data alchemy” from booking.com. Today we are struggling with multiple models, add more data and optimise the model, but what happens when we will change the collection of the data. Beside the data we have and the data we need we also have the part data you could create. An example for the question which data to collect and how booking.com asked the customer about their experience. The data scientists created magical models to analyse the sentiment of the feedback or we could easily change the way of asking:

The first day ended with a networking lounge in the bikini house. We had a great chat with a women from the berliner startup Blinkist and the CMO from runtastic.

I started the second day with a substantial breakfast and a talk of the Digital Growth Unleashed about “Visual storytelling” from eBusiness lab. I expected insights for data storytelling, but unfortunately the focus was more on Design and the zero moment of truth. After this talk I switched back to the predictive analytics world.

plists talked about “machine learning meets ad creation”. The starting point was to build a smart adcreator.

The learnings are dedicated time for experiments, test & iterate fast, early involvement of users, early commitment from management and usability.

The next talk from supercrunch focus on the development of a data science product. The message was “analytical products are different”. They used a hands on method for data design thinking with value proposition and business problems. With the canvas methodology they generated the users and problems to create benefits and features. The process of a MVP concerns feasibility checks/ data exploration, model prototyping and the product combined with engineer standards.

The talk about “dynamic forecasts” dived deep in the R programming to create dynamic linear models to predict the 400m olympic time.

The last presentation “you think what you consume” focused on the combination of Big Data and Neuromarketing. The representative of Neuro Flash integrated the embedded projector tensorflow.

To sum up, the conference was really inspired, at some talks I would have wished to be more specific.