How do I recognize 'good' clients?

As an insurance you have lots of new clients. Hows would it be if you had a risk assessment of the client already before contract start? Clients having a high risk could pay more than low-risk clients.

If you already have a risk-assessment it's based on rigid rules. New deeps to be detected and integrated in the software.

The rules are already contained in your contract data and damage reports. The question is:

How do you recognize in the data which rules should be used for customer categorization?

The good news is: you do not have to manually search the rules in the data. A deep learning software can analyze the data and, based on the insurance data, classify the potential contracts or predict the potential damage.

Details can be found here.