Wayfair Data Science Explains It All: Handling Imbalanced Data

Most machine learning algorithms are designed to train on balanced datasets. Resultantly, when our data is highly imbalanced, a typical model will have atrocious recall. In this video, Wayfair Senior Data Scientist Trent Woodbury explains the three most common ways of handling this imbalanced data problem. Trent Woodbury hails from Colorado, where he gained a BS in mathematics from Colorado State and a love of the outdoors which has turned him into an avid rock climber, hiker, and skier. While indoors, he enjoys reading anything from economic theory to cyberpunk mangas. Trent is spending this year at Wayfair’s Berlin office working on automating Wayfair's customer service through predictive modeling. Time stamps: 0:11 Imbalanced data 0:25 Binary classification examples 1:01 Handling imbalanced data 1:15 Undersampling 1:48 Oversampling 2:19 SMOTE - oversampling technique 6:57 Custom Loss Function 10:46 Overview