Dynamic Latent Class Model Averaging for Online Prediction

Abstract

We consider the problem of online prediction when it is uncertain what the best prediction model to use is. We develop a method called dynamic latent class model averaging, which combines a state‐space model for the parameters of each of the candidate models of the system with a Markov chain model for the best model. We propose a polychotomous regression model for the transition weights to assume that the probability of a change in time depends on the past through the values of the most recent time periods and spatial correlation among the regions. The evolution of the parameters in each submodel is defined by exponential forgetting. This structure allows the ‘correct’ model to vary over both time and regions. In contrast to existing methods, the proposed model naturally incorporates clustering and prediction analysis in a single unified framework. We develop an efficient Gibbs algorithm for computation, and we demonstrate the value of our framework on simulated experiments and on a real‐world problem: forecasting IBM's corporate revenue. Copyright © 2014 John Wiley & Sons, Ltd.

Dynamic Model Averaging in Economics and Finance with fDMA: A Package for R

“…Indeed, model averaging is a popular technique not only in the Bayesian setting. Various approach are developed, basing on different underlying theories [2,[72][73][74][75][76][77][78][79][80]. Recently, a very detailed surveys in model averaging in economics has been written [81][82][83], to which the interested reader is referred.…”

A Bayesian Model for Forecasting Hierarchically Structured Time Series

“…Large-scale organizations are often required to provide forecasts of a number of key performance metrics that measure attributes such as their financial health or growth potential (e.g., Yang, Hosking & Amemiya, 2014;Amemiya, Yang & Anikeev, 2014;Fan & Hyndman, 2008). These organizations are naturally structured in a hierarchical manner; they conduct business in various locations around the world and are split by divisions that are responsible for distinct products or lines of business.…”

Bayesian sequential parameter estimation with a Laplace type approximation

“…Nowadays, many statistical applications concern the analysis of streaming data for which fast algorithms are required e.g. dynamic classification (McCormick et al 2011) or online analysis of sensor data (Hill et al 2009, Yang et al 2015. A desirable property in these applications is that the inference result can be efficiently re-used or modified when another observation arrives.…”