Facebook Pixel

Applying AI to detect citizens misusing social benefits in Denmark

Horsens Kommune

Company description

Horsens municipality is a municipality in Region Midtjylland on the east coast of the Jutland peninsula in central Denmark. The municipality includes the island of Endelave, and covers an area of 524.21 km2. It has a population of 85.000.

The municipality was created in 1970 due to a kommunalreform ("Municipal Reform") that combined the city of Horsens with a number of existing parishes. The population in Horsens Municipality is one of the fastest growing in Denmark.

At our IT Department at Horsens Municipality we are mainly working on creating solutions to our employees and optimizing administrative tasks with machine learning, RPA and other technologies.

Project description

There are 400-500 cases of citizens misusing social benefits at Horsens Kommune each year. They come from anonymous claims by local citizens, typically neighbours. Around 50% of these claims are proven to be correct and then further legal action take place. However, the remaining 50% are false claims.

For each case or claim, our internal consultants are conducting an initial investigation which might take from 2 to 20 hours to research the case. It is a complex manual task to spot misconduct by looking into various databases. This might include, analysis of bank statements, loans data, significant purchases, flights abroad, social media analysis and others.

To improve efficiency, our department is creating Robotic Process Automation (RPA) robot to collect all the relevant information about the person once the claim is submitted via online form. This should drastically reduce time requirements for internal consultant when making initial investigation for each claim. We are launching the first version of the system from January and start to expand capabilities of RPA as we go along. The main task of RPA will be data aggregation and presentation.

The following step of RPA system will be to develop analytical capabilities to evaluate likelihood of any given case for being positive or negative based on data provided. To develop this function, we want to deploy Artificial Intelligence (AI) for data analysis and pattern recognition. We would like you to:

• Interview our internal consultants to understand what patterns they are using at the moment to spot misuse of social benefits.
• Help us decide what data needs to be collected and has the biggest predictive power.
• Develop machine-learning algorithm to evaluate accusations based on data provided.
• Construct feedback loop to continue improving algorithm prediction accuracy

The system once completed has potential and is expected to be implemented on a national level across all Kommunes in Denmark.

Student description

For this project we seek for a student or group of students within Master or Bachelor studies related to Data Science, Machine Learning, AI, Business Intelligence or similar.

We expect you to be self-driven, taking full responsibility for the project and leading it. In our team we have strong IT skillset so it is not required for you to be a software engineer or have programming background as long as you are skilled working with Data and understand agile process.

To support you through this project you will work directly with the Head of Development team at IT Department who will provide you with an access to required data as well as make introductions to other people working on this project. You will also have an office space at Horsens commune to come and work from here.

The applications will be reviewed on an ongoing basis and we will close the case once the right candidate is found. For any questions, please contact Kris at saikus@matchmythesis.com or +45 50280069

Applying AI to detect citizens misusing social benefits in Denmark | Match My Thesis
Dec 07, 2018
Bachelor, Master


Machine Learning




Public Sector