This paper develops an efficient approach to model and forecast time-series data with an unknown number of change-points. Using a conjugate prior and conditional on time-invariant parameters, the predictive density and the posterior distribution of the change-points have closed forms. The...