Carnegie Mellon University (CMU) and University of Porto team, won the 2017 IEEE International Conference on Data Mining (ICDM) Best Paper award for the paper “TensorCast: Forecasting with Context using Coupled Tensors“, a novel method that forecasts time-evolving networks like Twitter, for example. The conference will be held between November 18-21 in New Orleans, US.
The team members (l-r) Christos Faloutsos, Miguel Araújo and Pedro Manuel Pinto Ribeiro.
“TensorCast is able to forecast multiple co-evolving sequences, such as users buying products or user retweets. It takes into account side information (like demographics) and it scales up to millions of data points by carefully focusing on the few most active ones,” explain the authors Miguel Ramos de Araújo, Carnegie Mellon Portugal Program Computer Science Ph. D. Alumnus, INESC TEC Researcher, and Data Analyst at Feedzai; Pedro Manuel Pinto Ribeiro, Araújo’s Ph. D. co-adviser, Faculty Member at Faculdade de Ciências ofthe University of Porto, and INESC TEC researcher; and Christos Faloutsos, Araújo’s Ph. D. co-advisor and Faculty Member at CMU.
The ICDM is a top research conference in data mining and according to the team “despite all the data sources our systems have available, forecasts combining all this information are simultaneously very relevant and very difficult to create”.
The collaboration was enabled by the CMU Portugal Program, as both the Ph. D. funding and the travel support were instrumental to connect the two sides of the ocean.
The paper will be presented on November 19, at 2:00 pm.
Back in 2014, Miguel Araújo had already won a Best Student Paper Runner-Up Award at the Pacific Asia Knowledge Discovery and Data Mining 2014 (PAKDD) for the paper he co-authored with his advisors titled “Com2: Fast Automatic Discovery of Temporal (’Comet’) Communities”.
