ABSTRACT
There are many demands of time series analysis and forecasting, and in this chapter we will discuss some Bayesian approach to time series problems. We will start by considering time series modeling as a regression problem, with the design matrices parsed from the timestamp information. We will then explore the approaches to model temporal correlation using autoregressive components. We then show how these models extend into a wider (more general) class of State Space Model and Bayesian Structural Time series model (BSTS), also known as dynamic linear model (DLM), and specialized inference approach using Kalman Filter. The remainder of the Chapter will give some brief summary of model comparison as well as consideration of prior distribution for time series models.
