Predictive Analytics in Business Intelligence Systems via Gaussian Processes for Regression

Abstract

A Business Intelligence (BI) system employs tools from several areas of knowledge to deliver information that supports the decision making process. Throughout the present work, we aim to enhance the predictive stage of the BI system maintained by the Brazilian Federal Patrimony Department. The proposal is to use Gaussian Process for Regression (GPR) to model the intrinsic characteristics of the tax collection financial time series that is kept by this BI system, improving its error metrics. GPR natively returns a full statistical description of the estimated variable, which can be treated as a measure of confidence and also be used as a trigger to classify trusted and untrusted data. In our approach, a bidimensional dataset reshape model is used in order to take into account the multidimensional structure of the input data. The resulting algorithm, with GPR at its core, outperforms classical predictive schemes in this scenario such as financial indicators and artificial neural networks.