- Validis
Validis is a web based service which detects anomalies in accounting data in very much the same manner as a spell checker finds errors in a text document. The word itself is derived from a concatenation of Validate and Discover. The primary user base is among accounting professionals both within the financial community of corporate and SME business and across professional accounting practice. It provides a means to identify out of line data that is incomplete, invalid, inconsistent, or inaccurate, an analysis collectively tagged the ‘Four Is’.
The service, offered by [http://www.Validis.com Validis] is based on IP owned and patent protected by [http://www.future-route.com Future Route] .
Technology
Validis presents their service through a web-based interface. On the server side the system uses a number of sophisticated techniques to analyse data for quality problems.
Underlying the system is an abstract representation of accounting data, in the form of an ontology of accounting concepts. In an initial setup step the users accounts are automatically mapped into this representation by analysing the Chart of Accounts.
The system maintains an
Expert System in the form of a body of rules that describe valid, and invalid states for individual transactions, accounts, and other elements of the accounting ontology. Validis employs a specialised rule language developed in house, similar to Prolog and SQL, to create these rules. The rules are written in terms of elements of the accounting ontology, and therefore once a set of accounts has been mapped into it, the rules can immediately be applied to that set of accounts.Validis also uses
Information Theory to discover records in the data that are inconsistent with the typical behaviour exhibited in the data. By finding records which have excess information bits relative to the information bits of the individual field values in that record, Validis is able to identify unusual combinations of field values. It presents these unusual combinations to the user in the form of easily understood propositional rules.Numerical values in the data are also subjected to outlier analysis, individually, and agglomerated over time periods and over elements of the accounting ontology. Records with values that are outliers in absolute value terms, or that deviate from patterns identified in the data, are shown to the user.
Accounting data uploaded to Validis is also analysed and compared to the Benford Distribution (see
Benford's Law ), to check that the data has not been artificially generated or manipulated. The user is alerted if the data deviates from the expected distribution.Reporting
Output is presented in graphical form via a web interface. Underlying report data may be downloaded into excel for further analysis and manipulation. A presentable drill down analytical review may be manipulated across user selected time periods.
Timeline
While the parent business Future Route has been developing and deploying machine learning solutions since 2002. The service is due to enter beta release during March 2007 within a selected community of UK accountants, with full market availability soon thereafter. Initial deployments are compatible with SAGE (line 50), it is understood that further versions will be available shortly covering other versions of SAGE then MYOB, Intuit Quickbooks, Microsoft Dynamics GP and AX, Oracle, SAP, Sun Accounts,
An alpha test period operating throughout January and February 2007 resulted in user interface modifications and adaptations to bring the service in line with UK accounting industry expectations. Initial deployments are based on Sage Line 50 accounting data.
Following release of an initial library of business rules, capturing local legislative guidelines for accounts preparation, and typical accounting errors, such as missposts, omissions the service is expected to extend to include industry specific rule packs and advanced analytic functions and to support additional accounting packages throughout 2007.
External links
* [http://www.future-route.com Future Route]
* [http://www.validis.com Validis]
Wikimedia Foundation. 2010.