A semi-supervised tool for clustering accounting databases with applications to internal controls

A considerable body of literature attests to the significance of internal controls; however, little is known on how the clustering of accounting databases can function as an internal control procedure. To explore this issue further, this paper puts forward a semi-supervised tool that is based on self-organizing map and the IASB XBRL Taxonomy. The paper validates the proposed tool via a series of experiments on an accounting database provided by a shipping company. Empirical results suggest the tool can cluster accounting databases in homogeneous and well-separated clusters that can be interpreted within an accounting context. Further investigations reveal that the tool can compress a large number of similar transactions, and also provide information comparable to that of financial statements. The findings demonstrate that the tool can be applied to verify the processing of accounting transactions as well as to assess the accuracy of financial statements, and thus supplement internal controls.

► Self-organizing map combined with IFRS XBRL Taxonomy to perform clustering.
► Clustering of accounting databases as an internal control procedure.
► Clusters are homogeneous, and interpretable from an accounting perspective.
► Clustering serves to verify the processing of accounting transactions.