Automated explanation of financial data

We describe a methodology for explanation generation in financial knowledge-based systems. This offers the possibility to generate explanations and diagnostics automatically to support business decision tasks. The central goal is the identification of specific knowledge structures and reasoning methods required to construct computerized explanations from financial data and models. A multistep look-ahead algorithm is proposed that deals with so-called cancelling-out effects, which are a common phenomenon in financial data sets. Our method is an extension of the traditional variance decomposition in accounting. The method was tested on a case-study conducted for Statistics Netherlands involving the comparison of financial figures of firms in the Dutch retail branch.