Predicting videogame sales using Bayesian reasoning

Abstract

Financial data, such as data representing sales of a certain asset in a given economic sector, is intrinsically imbued with uncertainty, due to both the uncertainty of the financial market as well as the noise present in the data itself. In this report, probabilistic reasoning tools are applied to financial data concerning the sales in the videogames market. First, four different Bayesian Networks models are constructed either by explicitly defining the structure or by inferring it from the data. Then, for each model, parameters are learned from the available data. Finally, inference techniques are applied to perform probabilistic reasoning on the resulting model.

You can read the report here and see the code here.