Data Driven Model Learning for Engineers: With Applications to Univariate Time Series

The main goal of this comprehensive textbook is to cover the core techniques required to understand some of the basic and most popular model learning algorithms available for engineers, then illustrate their applicability directly with stationary time series. A multi-step approach is introduced for modeling time series which differs from the mainstream in the literature. Singular spectrum analysis of univariate time series, trend and seasonality modeling with least squares and residual analysis, and modeling with ARMA models are discussed in more detail.  As applications of data-driven model learning become widespread in society, engineers need to understand its underlying principles, then the skills to develop and use the resulting data-driven model learning solutions. After reading this book, the users will have acquired the background, the knowledge and confidence to (i) read other model learning textbooks more easily, (ii) use linear algebra and statistics for data analysis and modeling, (iii) explore other fields of applications where model learning from data plays a central role. Thanks to numerous illustrations and simulations, this textbook will appeal to undergraduate and graduate students who need a first course in data-driven model learning. It will also be useful for practitioners, thanks to the introduction of easy-to-implement recipes dedicated to stationary time series model learning. Only a basic familiarity with advanced calculus, linear algebra and statistics is assumed, making the material accessible to students at the advanced undergraduate level.

Table of Contents:
- 1. Basic Concepts of Time Series Modeling. - 2. Singular Spectrum Analysis of Univariate Time Series. - 3. Trend and Seasonality Model Learning with Least Squares. - 4. Least Squares Estimators and Residuals Analysis. - 5. Residuals Modeling with AR and ARMA Representations. - 6. A Last Illustration to Conclude.

About the Author :
Guillaume Mercère received the M.S. degree in electrical engineering in 2001, the Ph.D. degree in automatic control (Lille University) in 2004 and the "Habilitation à diriger des Recherches" in 2012. Since September 2005, he has been an Associate Professor at Poitiers University, Poitiers, France, and a member of the Automatic Control and Electrical Engineering Laboratory of Poitiers. He was chair of the Electrical Energy Optimization and Control Department, Poitiers National School of Engineering, between 2010 and 2015. He was the co-leader of the French Technical Committee on System Identification between 2008 and 2014, then the chair of the IEEE CSS Technical Committee on System Identification and Adaptive Control between 2016 and 2019. He is currently an Associate Editor on the IEEE CSS Conference Editorial Board. He is the co-author of more than 80 international conference and journal papers. He has held visiting appointments at the University of Iceland, Nova SoutheasternUniversity in Florida (USA) and Politecnico di Milano in Italy. His main research interests include model learning and system identification theory, estimation theory, optimization theory, subspace-based identification for 1D and nD models, gray box and linear parameter varying system identification with a specific attention to state space models. His current activities focus on heat transfer, flexible and cable driven manipulators, aeronautics, vehicle tire/road interactions and image processing.