Recently, several approaches have been proposed for learning Bayesian networks. These methods combine prior knowledge with data to produce one or more Bayesian networks. The resulting networks can be used for prediction, or, in special cases, to infer causal relationships among variables. In particular, in financial market applications, the resulting model will be able to help analysts predict future market direction, understand critical market parameters, and make investment decisions. By using Bayesian network techniques, this thesis aims to discover network representations for analysis earnings from the viewpoint of fundamental analysis. Based on Bayesian network learning algorithms, the causes (financial ratios) and effect (earning per share) network representations are generated for encoding uncertain beliefs and drawing inferences from such representations. The empirical results show that the accuracy is about 80%.