Predictive Analytics in the Public Sector: Using Data Mining to Assist Better Target Selection for Audit

Abstract

Revenue, the Irish Tax and Customs Authority, has been developing the use of data mining techniques as part of a process of putting analytics at the core of its business processes. Recent data mining projects, which have been piloted successfully, have developed predictive models to assist in the better targeting of taxpayers for possible non‑compliance/ tax evasion, and liquidation. The models aim, for example, to predict the likelihood of a case yielding in the event of an intervention, such as an audit. Evaluation cases have been worked in the field and the hit rate was approximately 75%. In addition, all audits completed by Revenue in the year after the models had been created were assessed using the model probability to yield score, and a significant correlation exists between the expected and actual outcome of the audits. The models are now being developed further, and are in full production in 2011. Critical factors for model success include rigorous statistical analyses, good data quality, software, teamwork, timing, resources and consistent case profiling/ treatments. The models are developed using SAS Enterprise Miner and SAS Enterprise Guide. This work is a good example of the applicability of tools developed for one purpose (e.g. Credit Scoring for Banking and Insurance) having multiple other potential applications. This paper shows how the application of advanced analytics can add value to the work of Tax and Customs authorities, by leveraging existing data in a robust and flexible way to reduce costs by better targeting cases for interventions. Analytics can thus greatly support the business to make better‑informed decisions.