Max Izatt is lead data scientist in the Insight & Analytics Group at Hitachi Consulting. Trained as a theoretical chemical physicist at MIT and Chicago, Max began his career with Bell Laboratories and Sandia National Laboratories in the Laser Projects Division of the Pulsed Power Research Group, where he worked on laser-switching circuitry for space-based directed-energy platforms. In the private sector, Max pivoted to fixed-income algorithmic trading at Salomon Brothers, where he was lead quantitative programmer in the Tokyo financial group.
He founded CentraLytics Corporation at the Chicago Mercantile Exchange in 2002, which developed distributed-computing and collaboration features that coupled the desktop with the Microsoft enterprise business-intelligence stack for near-real-time high-frequency trading applications. CentraLytics was successfully acquired within the Microsoft footprint in 2014. At Hitachi, Max works on queuing, quality-of-service, and credit-scoring topics for communications, media, and entertainment firms; and operations-research topics for manufacturing, retailing and distribution firms.
A data-driven/machine-learning approach to real-time network monitoring and fault detection
We will discuss the statistical thermodynamics, information theory, and data science of anomaly detection at line rate for PIM/IGMP packet streams as applied to near-real-time quality-of-service (QoS) monitoring, identification, remediation, and repair. We will review a case study from the October, 2016 time frame in which machine-learning algorithms detected network anomalies, which allowed a multi-system operator (MSO) to proactively groom the network in anticipation of a highly-subscribed high-definition television broadcast. QoS was statistically inferred from real-time and after-action call-center statistics that show a statistically-significant improvement in the affected markets versus the national baseline. Using operations-research techniques, we quantify the benefit of the data-science platform to the MSO’s business.