Model Multinomial Bayesian Network pada Data Simulasi Curah Hujan

Abstract

Bayesian Networks is one of simple Probabilistic Graphical Models are built from theory of bayes probability and graph theory. Probability theory Is directly related to data while graph theory directly related to the form representation to be obtained. Multinomial Bayesian Network method is one method that involves the influence of spatial linkages suggest a link between rainfall observation stations. The objective of this study was seek the result of the model probabilistic a graph Multinomial Bayesian Network and apply it in forecasting with Oldeman classification based on one or two rainfall stations are known. This research uses simulated data for 14 stations respectively each 300 sets of data. The data generated is normal distribution of data based on parameters that have been determined and classified using the classification Oldeman. Bayesian Network structure constructed using the K2 algorithm. Markov chain transition matrix is formed based on the Bayesian of the nodes are directional. Model of Multinomial Bayesian Network was established based on Markov transition matrices. The result of probability model can predict the probability of rainfall in some stations based on one or two rainfall stations are known, which is a model graph with 14 nodes and 13 arcs.