Probabilistic Inference Based on LS-Method Modifications in Decision Making Problems

Abstract

The article provides information on the modification of the Lauritzen-Spiegelhalter (LS) method for constructing a probabilistic inference in the Bayesian network. Modification of the method consists in a new way of filling tables of conditional probabilities. The method consists of two stages: the first stage – the construction of a combined tree, the second stage – the construction of the distribution algorithm. At the first stage, the construction of a combined tree is performed by clicking on the primary structure of the network and filling the vertices of this tree with tables of conditional probabilities of the network. The second stage of the LS-method is the refinement of the distribution algorithm. The modified method allows us to more accurately calculate a posteriori probabilities and build a probabilistic inference. The example of evaluation of decision-making options for the construction of solar and wind power plants is considered. The features of the application of the method are analyzed.