GBM: stories from the trenches
Exploiting R to its maximum potential is often seen as a difficult task: who thought we could use R and GBMs on a server with several hundred cores to its maximum? This talk will present several case studies where combining GBMs and using R performs extremely well in both production and R&D situations to answer practical business use cases. Some of the examples include very peculiar usages of GBMs.
Examples in production include using 1 million monotonic GBMs for delivery optimization throughout a whole country, and showing optimality of flavor atomic coordinates.
Examples out of production include performing backtesting on years of data to assess forecasting accuracy to minimize human expenses, and optimizing bank bonds.
SAP BI / Data Science Consultant, Planeum
Damien obtained a MSc by analyzing the impact of reporting visuals on decision-making in businesses for diagnosing and forecasting. Since then, he worked as an entrepreneur then joined Planeum to push the innovation a step ahead, combining IT, BI and Data Science knowledge.
He is now focusing on combining Design and Data Science skills, benchmarking machine learning tools, and human psychology knowledge. He aids to the development of R packages such as LightGBM (collaborator), and to the performance of Data Science tools such as xgboost (contributor).