Event Timing: May 16th, 14:30
Organization: DCC-FCUP

Abstract: The volume and incoming speed of data have increased tremendously over the past years. Data frequently arrive continuously over time in the form of streams, rather than forming a single static data set. Therefore, data stream learning, which is able to learn incoming data upon arrival, is an increasingly important approach to extract knowledge from data. Data stream learning is a challenging task, because the underlying probability distribution of the problem is typically not static, but suffers changes over time. Such challenge is exacerbated by the fact that the data distributions are often skewed. In classification problems, this means that the number of examples of a given class of interest is small compared to others, making adaptation to changes affecting such class difficult. In this talk, I will discuss online oversampling approaches that can be used to help tackling class imbalance in data stream learning.