Spatial Bayesian Network

One of the important characteristics of Bayesian network is that it can intuitively model the dependency among numerous variables. However, as the network becomes large, containing several nodes and edges, the computational complexity of Bayesian network (BN) analysis increases at a large extent. Now, in many cases of spatial time series prediction, it is necessary to take into account the influences of variables from large number of spatially distributed locations, which becomes almost intractable for a standard BN model. In this chapter, we introduce SpaBN, a spatial extension of Bayesian network, which has intrinsic capability of efficiently modeling inter-dependency among very large number of spatially distributed variables. SpaBN treats the influences of same variable from spatially distributed locations to be a combined influence from a single composite node. Replacement of such standard nodes with equivalent single composite node drastically reduces both structural complexity and algorithmic complexity in SpaBN analysis. The performance of SpaBN has been evaluated in comparison with a number of conventional statistical and state-of-the-art space-time prediction models, with respect to case studies on climatological and hydrological time series prediction. Experimental result demonstrates the effectiveness of SpaBN in spatial time series prediction under profusion of influencing variables.