Master students' Big-Data company comes 1st in Kaggle competition
The L2F team: Francesco Palma, Thomas Boys, Aldo Podestà, Wallyson Lemes (credit: Aldo Podestà)
A team of Mathematics Master students from EPFL have founded a Big Data consultancy that has now won a Kaggle competition.
Encouraged by their performance in the 2017 Data Mining Cup where they came 13th, the students, Aldo Podestà, Wallyson Lemes de Oliveira, and Francesco Palma, with Thomas Boys who recently graduated in Math from EPFL and Paolo Tournon (HEC- Paris), went on to found L2F, “a lean, data-driven and performance-enhancement task force”.
The company provides Big Data consultancy services and uses topological and statistical modeling to “understand, forecast, and finally extract value from complex, company-specific data sets, throughout the whole spectrum of corporate divisions”. These include marketing and sales, customer relations, human resources and other areas where Big Data problems may occur.
As an initial project, the L2F team took part in a competition on Kaggle, which is the world’s largest data-science community. The competition’s objective was to predict the durations of taxi trips in New York. The EPFL students came first out 1,257 competing teams, making an extremely accurate prediction on a dataset with more than 600,000 trips.
“At the beginning of our experience in the field of Big Data we found a highly competitive environment and we immediately realized that would have needed to bring innovation to outperform existing approaches,” says Aldo Podestà. “For this reason, we started exploring concepts coming from pure mathematics such as topology and graph theory, which seemed to better describe the ‘shape’ of the data compared to other, merely statistical methodologies. After our winning performance in the Kaggle competition, we believe, even more than before, that such a holistic attitude could potentially open up new horizons when it comes to complex predictive analysis.”