Open PhD positions at University of Antwerp in Data Mining


Explainable AI

Most of the recent AI applications are found in predictive data mining: predicting some target variable (interest in an ad, fraud, credit default, churn, object in an image, sentiment of document) based on various inputs. A major open research issue these days is explaining why an AI model made a certain prediction: why was this company flagged as likely fraudulent, why was the car in this image predicted to be total loss, why was I shown this ad? These are hard questions to answer, mainly due to the complexity of the prediction models. This complexity arises from either the technique (for example deep learning), or from the huge-dimensional input data (for example Facebook like or payment data). In this research we develop and apply new techniques that explain complex prediction models. This research continues and can build on our initial work on evidence counterfactual.

Read more: https://www.computerweekly.com/news/252457364/Explainable-AI-How-and-why...
Our own work: http://pages.stern.nyu.edu/~fprovost/Papers/MartensProvost_Explaining.pdf
https://www.sciencedirect.com/science/article/abs/pii/S0377221706011878

Fraud detection with behavioral data

The current fight against fraud is of fundamental importance, given the impact of fraud on economy and society. To illustrate, the Belgian government loses about 25 billion euro annually as a result of tax fraud, which corresponds to 6% of its gdp. In the EU, this figure rises to an astonishing 1 trillion euro. Current research in data mining to find fraud patterns in large volumes of data is still in its infancy, and a wide range of data types are awaiting to be researched. For example behavioral data on actions and interactions of persons and companies: think for example of data persons and harbors that containers have been handled by to assess customs fraud risk, or invoices sent and received by companies to predict VAT fraud, or detecting patterns in the network of persons owning shares of certain companies. In this research area, we will develop new techniques, tailored to such data, and the validation thereof in various tax fraud areas, based on previous studies done with the Belgian tax authorities.

Read more: https://www.information-age.com/fraud-data-mining-123481177/
Our own work: https://www.researchgate.net/publication/266660335_Corporate_residence_f...
https://www.gva.be/cnt/dmf20190303_04224894/doctorandus-kan-fraude-18-ke...

Profile

- Master’s degree in business engineering, civil engineering, mathematics, physics, or a closely related discipline with relevant experience in mathematics, statistics, and/or data science.
- Strong affinity for creative data analysis, a broad interest in data science, numerical problem-solving and analytics.
- Enthusiast for doing research with positive mindset, critical thinker.
- Experience in programming (eg, Python, Matlab).

What do we offer

- PhD position (3-4 years) in the Applied Data Mining research group of University of Antwerp.
- Attractive salary that is competitive with the private industry.
- Young, dynamic and inspiring environment with the possibility to engage in fun activities such as after work drinks, paintball, running sessions, group weekends, etc.
- Ability to work on relevant, state-of-the-art research with Belgian companies.
- Possibility for an international research stay abroad.
- Office with great view on the beautiful city of Antwerp

How to apply

Please send your application to Prof. David Martens (david.martens@uantwerpen.be)

Author: 
yanou ramon