Time Series Statistical Learning and Control

Catalog description

This course covers applications in industrial engineering such as statistical process control and supply chain modeling (controlling the "bullwhip" effect). Both polynomial and state-space models are discussed. The former includes ARIMA and Transfer Function (Box-Jenkins) models, while the later include Kalman filtering, smoothing, and optimal stochastic control. While the bulk of the course deals with equidistant observations over time, Time Series models for irregular observations over time are dealt with by introducing continuous time AR processes. Multivariate time series models are introduced next, emphasizing their interpretation by means of their graphical (network) representation and the use of Dynamic Mode Decomposition techniques. The course includes a discussion of Recurrent Neural Networks for forecasting, and concludes with an introduction to Spatio-temporal data and to the Topological data analysis of time series.