Predictive analytics on evolving data streams anticipating and adapting to changes in known and unknown contexts

Ever increasing volumes of sensor readings, transactional records, web data and event logs call for next generation of big data mining technology providing effective and efficient tools for making use of the streaming data. Predictive analytics on data streams is actively studied in research communities and used in the real-world applications that in turn put in the spotlight several important challenges to be addressed. In this talk I will focus on the challenges of dealing with evolving data streams. In dynamically changing and nonstationary environments, the data distribution can change over time. When such changes can be anticipated and modeled explicitly, we can design context-aware predictive models. When such changes in underlying data distribution over time are unexpected, we deal with the so-called problem of concept drift. I will highlight some of the recent developments in the proactive handling of concept drift and link them to research in context-aware predictive modeling. I will also share some of the insights we gained through the performed case studies in the domains of web analytics, stress analytics, and food sales analytics.