Dynamic Bayesian approach to forecasting
Adelina Tang · 2010 Sixth International Conference on Natural Computation · 2010
Bayesian belief propagation is flexible and highly adaptable in machine learning and artificial intelligence methodologies. Coupled with a time element, the Dynamic Bayesian approach has shown promise in forecasting applications. A methodology consisting of beliefs propagated through the TAN-Pearl network and computed for every time slice is proposed to this end. Benchmark comparisons indicate encouraging results.