Are you more than familiar with programming and looking for a challenging internship and/or graduation project, then this is the spot for you!
Benchmark studies estimate that around 5-15% of incoming insurance claims are unlegitemate to some degree. As an insurane company who handles >1000 claims a week, it is essential that we process claims fast for honest customers, and on the other hand double check suspicious claims extensively. We have succesfully shown that anomaly detection can help us intercept suspicious claims more efficiently. Literature shows there are many different options for anomaly detection amongst which neural networks seem to be very promising.Project Goal
Design and build an unsupervised autoencoder neural network for anomaly detection. The model can be build in python or R. In addition, a comparative study of interpretation methods (e.g. LIME) will really help us explain predictions to the business. The assignment means you will do a fair share in coding and tweaking but also leaves plenty of room for literature research. Specific scope can be further determined as you see fit, weighting in how practical or theoretical your assignment needs to be (e.g. to graduate).We are looking for
- A neural network/ML expert or enthusiast looking for a graduation project or internship
- You are currently in your master Data Science, Data Analytis or an equivalent study
- You have programming experience
- Can start as soon as possible
- An assigment for a minimum of 3 months at Aegon NL in the Hague (3-5 days a week)
- Opportunity to gain experience in creating real data science applications for business
- Working in a young team of data analysts
- A monthly internship allowance
If you are interested or have ay questions regarding the project, please contact firstname.lastname@example.org with a short introduction.
- Den Haag