The Operationalization of Machine Learning Models

PROCESS OPTIMIZATION

Efficiency gains through intelligent automation

The intelligent use of advanced analytics, automation and machine learning is becoming increasingly important for banks in fraud prevention, especially in the credit risk assessment process. This approach enables analysis of huge amounts of data more efficiently than ever before, unearthing of suspicious patterns and thus, identifying potential risks at an early stage. This, in turn, reduces effort, increases efficiency and lowers costs. However, the operationalization of ML models, in particular, presents companies with major technical challenges. This case study on automated earnings statement control in the credit risk assessment process shows how the above approach can be optimally achieved by using ACTICO Platform – by the combination of expert judgement rule models and machine learning algorithms.

CHALLENGE

Digital Processes & Machine Learning

The high number of manual checks in the credit risk assessment process ties up human resources which could be better utilized for other, higher-value tasks.

The introduction of digital processes, Advanced Analytics (AA) and Machine Learning could reduce the number of manual checks and thus, significantly increase efficiency.

SOLUTION

Operationalization of machine learning models

• Development of a statistical forecasting model based on machine learning algorithms, which evaluates the probability of fraud in a credit application and thus, enables a targeted management of questionable applications.
• Operationalization of the forecast model.

RESULT

Business Value

• Efficiency gains: the number of manual checks on earnings certificates was reduced by 80%.
• Through this process automation, VW FS achieved an annual savings of over 1 million euros.

OUTLOOK

Perfect Collaboration

This case study demonstrates how significant the efficiency gains can be through the combined use of ML processes and intelligent automation. The ACTICO Platform offers an effective solution to overcome the hurdle of operationalizing ML models. It brings data scientists together with the IT department, creates a bridge between the programming languages, Python and Java, and ensures smooth functionality. This reveals a great potential for companies in all areas where ML models are used – regardless of industry, size or department.