Clustered data is all around us. The most common example is longitudinal clustering, where each individual instance of a phenomena you wish to model has multiple associated measurements (e.g., modeling math test scores as a function of sleep factors when you have multiple measurements per student). Another common example is clustering due to a categorical variable (e.g., clusters representing the specific math teacher of a group of students). Clustering can also be hierarchical (e.g., a student cluster contained within a teacher cluster, which is itself contained within a school cluster). When modeling clustered data, you must account for any idiosyncrasies and nonnegligible random effects by cluster.
The best way to attack this kind of data? Mixed effects models. Inspired by the models we have been building for clients, Manifold has developed mixed effects random forests (MERF)—an open source implementation package in Python.
Sourav Dey explains how the MERF model marries the world of classical mixed effect modeling with modern machine learning algorithms and shows how it can be extended to be used with other advanced modeling techniques like gradient boosting machines and deep learning. He also walks you through example use cases and demonstrates MERF performance on synthetic and real data.
Sourav Dey is CTO at Manifold, an artificial intelligence engineering services firm with offices in Boston and Silicon Valley. Previously, Sourav led teams building data products across the technology stack, from smart thermostats and security cams at Google/Nest to power grid forecasting at AutoGrid to wireless communication chips at Qualcomm. He holds patents for his work, has been published in several IEEE journals, and has won numerous awards. He holds PhD, MS, and BS degrees in electrical engineering and computer science from MIT.
Comments on this page are now closed.
©2019, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com