Applying A.I. to Diverse Transactional Data For Better Decision Making

Transactional data comes from very diverse sources, and is categorical in nature, often with attributes such as a person’s gender, marital status, hometown, career, or the types of movies they like. At MediaGamma, a UCL spin-out, they have developed an Audience Decision Engine (ADE), that can teach machines to make various predictions and then subsequently make optimal actions for terabytes transaction data.

Dr. Jun Wang is Co-founder and Chief Scientist at MediaGamma, as well as Associate Professor in Computer Science, and Director of MSc Web Science & Big Data Analytics, at UCL. In his talk at the Deep Learning in Finance Summit, Jun will look at AI solutions for transactional data generated from our social and commercial world, such as how supervised and unsupervised learning can be used to decipher correlations in different attributes, and use them to make various accurate forecasts and how this knowledge can be transferred for predictive tasks and optimal decision-making.

I asked Jun a few questions to learn more about advancements in the deep learning field, as well as the risks and challenges we face when applying AI technologies in the finance sector. 

What have been the leading factors enabling recent advancements and uptake of deep learning?
Deep learning has had different names in the past. It is essentially neural networks, which can be traced back to Rosenblatt's single layer networks called Perceptrons in 1958, whereas convolutional neural networks with multiple hidden layers (which is why it is called deep learning) were invented as early as 1989. The core underlying learning algorithm, back-propagation, has been re-invented quite a few times in history by different people.

In my view, it's a natural progression in science and technology where as data becomes larger and the problem becomes more sophisticated, we need more expressive mathematical models to decipher the underlying structure of the data and extract insights from it.

Deep learning as a family of machine learning methods fulfils that need. For instance, it allows a multi-class output and that can be scaled to thousands class or even millions; some time dependent neural networks allow sequence to sequence predictions. Also, making the neural networks deep would uncover high-level feature representation etc.

In addition, media exposure on recent progress of image recognition, speech recognition and machine translation and a few recent acquisitions also increases the public awareness and interest of the subject.

What is essential to continuing this progress?
An interesting observation is that the research that used to be handled mainly by universities and research institutes becomes more open and there are many joint efforts from both academia and industry to advance the field. Scientific findings and latest test results (quite often along with open source codes) are widely available in open platforms such as and Github in no time.

To keep the momentum going, we definitely need such an open ecosystem.

What area of finance do you feel will be most disrupted by AI?
Finance essentially addresses a resource allocation problem. Many areas will benefit from the automation and better optimisation resulted from AI, but I feel retail banking would be disrupted more quickly, for instance fraud detection, user acquisition, CRM and robo advisors to name a few.

What risks and challenges do we face when applying AI to financial services?
Risks: regulations.

Challenges: open data. Traditionally in terms of sharing data, financial sectors are very closed. AI needs to work with large scale, aggregated data.

What advancements in deep learning do you feel will impact the financial sector the most in the next 5 years?
I think, firstly, time series modelling; the ability of making use of and innovating time dependent models such as Recurrent Neural Networks. There are some impressive results from those models in other domains. It would be very interesting to see the applications in finance. Secondly, I also feel that the ability of shifting from predictive analytics to automatic decision making will make a huge impact on the financial sector.

Jun Wang will be speaking at the Deep Learning in Finance Summit on 23 September, which is taking place alongside our flagship Deep Learning Summit in London. Discounted passes are available until Friday 29 July! Previous events have sold out, so book early to avoid disappointment. For more information and to register, please visit the website here.

We are holding summits focused on AI, Deep Learning and Machine Intelligence in London, Amsterdam, Singapore, Hong Kong, New York, San Francisco and Boston - see all upcoming summits here.

Big Data Machine Learning Deep Learning AI Deep Learning Summit Deep Learning in Finance Summit FinTech Deep Learning Algorithms


    As Featured In


    Partners & Attendees

    Ibm watson health 3.001
    Rbc research.001
    Mit tech review.001
    Kd nuggets.001
    Maluuba 2017.001
    This website uses cookies to ensure you get the best experience. Learn more