Predictive Analytics Competence Center Predictive Analytics

Efficient and targeted risk management through machine learning

Predictive Analytics is an approach used in data analysis. It uses historical data to predict future events. The historical data are used to create mathematical models, which are then applied to current data. Machine learning is often mentioned in this context. Supervised machine learning techniques are frequently used. This means that the models are fed with input and output values, also known as training sets. Thus they “know” the data or correct answers. Predictive analytics methods require a large stock of historical data to learn from. The models then apply the acquired knowledge to new data. Thus, the models are able to predict what will happen next, or which actions will lead to an optimal result, respectively.

The Predictive Analytics Competence Centre (PACC) is expected to make, by continuous further development of the workflows, a significant contribution to supporting the strategy of the Federal Ministry of Finance in terms of setting up a modern risk management structure, increasing efficiency in auditing and anti-fraud activities, and tax collection. By apportionment of the tasks to 4 subject areas, each topic can be optimally addressed. The 4 subject areas are:

• Predictive Analytics (PA)
• Advanced Analytics (AA)
• Tax Analytics (TA)
• Customs Analytics (CA)

Predictive Analytics

This focus covers all the areas that have to do with statistical and mathematical methods and models – also known under the keywords “artificial intelligence” or “machine learning”. This ranges from development through maintenance to optimisation of these models. It also includes pre-processing of the data provided for analysis.

Advanced Analytics

This division deals with data analyses that are not related to Predictive Analytics methods. Here, all the issues pertaining to audit support for the tax and (G)PLA auditors are encapsulated ((G)PLA = (joint) audit of wage-related taxes) – i.e. all the topics relating to automation using ACL and ACL Mentor, as well as topics concerning cash registers and the implementation of the requirements of the Standard Audit File–Tax (SAF-T).

Tax Analytics

In the Tax Analytics division, all the areas are going to be consolidated that deal with fiscal analyses or tasks, respectively. This applies to both business assessment and general assessment, as well as to wage-related taxes. From the Tax Analytics unit, project managers for the (G)PLA and tax audit as well as other fiscal projects will be selected.

Customs Analytics

This pillar brings together all the customs-specific tasks. This includes the preparation of risk profiles, the new or further development of customs systems up to the support of customs colleagues in the risk assessments of consignments, and the implementation of international anti-fraud projects.

The main tasks of the PACC are the implementation of projects in the area of general assessment or of a real-time risk assessment, respectively, centralised audit case selection, establishment and expansion of automated examinations in the context of tax audits, company start-ups, analyses in the area of VAT missing-trader fraud, and support services in the area of data management and combating fraud.

Through use of Predictive Analytics, the PACC has been able to make a significant contribution to the attainment of a surplus in terms of tax refund claims in recent years. It may be assumed that this number will continue to rise over the next years as the Predictive Analytics methodology is being expanded. However, it must be emphasised here strongly that the imposition of a surplus of tax refund claims is always only and exclusively carried out by the auditing bodies of the Federal Ministry of Finance. Here, the PACC has the function of a service department that provides information on high-risk cases.