Machine Learning for Urban Computing

This chapter discusses the key techniques in machine learning (ML) and their use in various urban computing (UC) scenarios. It introduces the main concepts and key techniques of ML, followed by relevant issues while designing any ML study. ML methods can adapt to dynamically changing environments, which is very important in UC, as cities are constantly in motion. The capabilities of ML have increased in the last decades owing to increased abilities of computers to store and process large volumes of data, as well as because of theoretical advances. Artificial neural networks are cognitively inspired models and they have a lot of applications in ML. Bayesian approaches aim to model our state of knowledge, where prior knowledge is modified by empirical evidence, which arrives in the form of observations. Bayesian approaches are especially useful in modelling dynamic systems. Graphical models are large class of Bayesian approaches including Bayesian networks, Markov networks, random fields, and more.