Bayesian Learning in Share Selection

The return and various attributes from the McGreggor database were used to estimate the return that will be achieved by different companies in the Johannesburg Stock Exchange (JSE). None of the attributes were significantly correlated to the return. The Pearl algorithm in Bayesian belief networks induces a belief network from data. With a solid grounding in probability theory, the Pearl algorithm allows belief updating by propagating likelihoods of leaf nodes (variables) and the prior probabilities. The Pearl algorithm was originally developed for binary nodes, but this is generalized to more states. The results of this model were then compared to a Linear Regression model. The Bayesian belief network outperformed stepwise linear regression