Avi Mandelbaum (Technion) - Data-Based Service Networks: A Research Framework for Asymptotic Inference, Analysis and Control of Service Systems Thursday, September 27, 2012 at 3:00pm Frank H. T. Rhodes Hall, 253 The lecture is on Service Systems, pertaining to telephone call centers, hospitals, public services, banks, airports, supermarkets and more. The focus is on operational performance - e.g. customer flow and system congestion (with operational characteristics serving also as surrogates for financial, psychological and clinical performance). And my mathematical framework is asymptotic queueing theory, specifically fluid approximations and their diffusion refinements. Queueing theory is ideally suitable to capture the operational tradeoff that is at the core of any service - quality vs. efficiency. Asymptotic analysis has the potential to accommodate service characteristics which are otherwise intractable. Two cases in point are the Erlang-A and Erlang-R models: the first has become the most prevalent call center model, and the second applies to emergency departments - both are simple models that valuably portray complex realities. This raises several fundamental research questions: why are such simple models robust; what are the limitations of their robustness; and which alternative models would overcome the limitations. I shall address these questions within my asymptotic framework, developing models directly from data and validating their value against actual service systems. (This is in contrast to prevalent practice where models are typically remote from data, and asymptotic approximations are validated for accuracy against their originating mathematical models.) The ultimate goal is a platform for automatic creation of data-based asymptotic models, which will be universally accessible for applications by researchers, students and ultimately practitioners.