How will she spend the prize money? “It’s fifteen thousand euros,” Ines Wilms (30) says, sounding a little abashed. It’s a lot of money. She could use it to buy a heavy-duty computer, one which has 32 processor cores, eight times as powerful as a modern laptop. It costs five thousand euros, but it provides the kind of computing power needed for her research. She could use the rest of the money to visit colleagues in the US, England and Australia. “In addition to communicating via e-mail, it’s sometimes good to see each other and to be near each other to work on a project for a few weeks.”
Her roots are in Dilbeek, Belgium, a municipality neighbouring Brussels, less than twenty kilometres from Leuven. That’s where her academic training took place: she studied, received her PhD and held a postdoc position there. She became an assistant professor at the UM School of Business and Economics a year ago.
Wilms knows everything there is to know about big data. More specifically: about analysing big data. Even more specifically: about statistical methods for analysing big data. She’s currently working on a project commissioned by the European Commission to find out how much steel EU countries import from outside the EU and which type of import fraud occurs most commonly. Asian countries in particular have a tendency to dump steel on the European market at rock-bottom prices. The EU imposes high anti-dumping duties on steel, which are evaded in two ways: by using another country as an intermediate destination or by tampering with the product code, making the steel look like a different type of steel.
Wilms designed a statistical method (a network algorithm) that can detect anomalies in the vast amount of global import data from all countries. The software searches for detours to find intermediate destinations, for example, or detects surges in anomalous types of steel on the market. “With this type of research, the trick is to translate the request into an appropriate method.”
Such requests can also come from industry. Wilms once wrote software for food company Danone that automatically calculates the demand for yoghurt in supermarkets by week. Her program took into account all kinds of variables, such as sales promotions. “Consumers who make use of the ‘buy 3, get 1 free’ offer one week will buy less yoghurt the next week.”
Those kinds of variables are the strength of current statistical methods, says Wilms. “You can unleash an infinite number of variables on data. For example, I’ve done research for the European Central Bank, which wants to forecast the gross domestic product per country. The software takes into account inflation, interest rates, production by sector, the stock market index, and so on. Economists love their variables, but they have to be relevant.”