Quantitative investment methods have gained foothold in the financial world in the last ten years. This paper shows how Bayesian Networks can be used to create a computerized stock-picking model.
By using historical
data for 14 different economic relevant variables the model is
designed to give trading recommendations (buy or sell) for the
different companies included in a given dataset. The model has
a hitrate of 60% and it generates an average return of 15.1% in
each of the five investment periods tested. The model is found
to give a significant higher return than the mean value of
randomly generated portfolios and can therefore be said to
posses skills when it comes to stock picking.
