Time Series Modeling and Forecasting

This chapter presents temporal model-based approaches to analytics in dynamic, fast-changing environments such as the financial market. Temporal approaches, as opposed to static-or snapshot-based approaches, handle time formally, in that random variables in a temporal model change their values as time changes to form a time series. Temporal models are therefore appropriate for modeling the stochastic nature of a dynamic environment such as the financial market. We present the Kalman Filter (KF) technique for estimating the state of a dynamic system where transition from one state to another is assumed to be linear. We then discuss nonlinear transition via the Extended Kalman Filter (EKF).