An Assessment of Markov Renewal Models in Forecasting International Affairs

A major difficulty for designers of systems to forecast international affairs has been to allow the use of both observable data and subjective estimates. Under DARPA sponsorship, a novel approach has been developed based upon a Bayesian stochastic model. This approach has been developed and demonstrated in a preliminary fashion by other DARPA contractors. The present report provides a brief appraisal of the approach, offers preliminary suggestions for modifications, and suggests candidate areas of application. The treatment of time and nonstationary is explored in some detail, and modifications are suggested which could lessen the need to reassess model parameters each time a change occurs in an underlying political process.