Abstract: Anomaly detection is a common set of techniques used to address common problems faced by many organizations. Approaches are well-understood and implementation strategies are mature and varied. Despite the availability of computing power and the proliferation of expertise and effective tools, however, challenging problems remain. Black swan events exist; subjectivity is the ultimate definition of true/false. In this talk, we’ll discuss some strategies for incorporating human judgment into the prediction-detection loop in order to improve accuracy, produce more true positives and fewer false positives, and improve user satisfaction.
Bio: Andrew is the head of data science at Bigeye, a data observability company. Prior to joining Bigeye, Andrew built ML-powered tools for Citi and (as a consultant) a range of top consumer banks; he specialized in pricing and underwriting problems. In his free time, Andrew enjoys cooking, travel, and using his TVR Chimaera to escape New York.
