So I am tackling this problem statement in Insurance domain(fairly new to it). We have historical insurance claim data e.g. FNOL (first notification of Loss), Damage description & report, description of event, date time, location of loss, policy details, Adjuster Assigned to case, Time taken to complete the claim, Final Payout and Adjuster Meta data and other many more features.
Now the problem statement is - For each claim we have to identify the best possible adjuster or adjusters(one claim can have multiple adjusters base on the exposure) and to minimize the claims adjustment expense, reduce claims duration, Improve accuracy and Customer Satisfaction rate. Adjuster skill identification and the adjuster identification is the final goal.
The existing system is mainly rule base without any machine learning. For existing claims they have categories( like fast track, complex & fraud). The idea is not to replicate the current assignment process but to segment the claims to achieve the main objective of assigning best adjuster or adjusters to a claim.
One approach we are thinking is to make bucket of claims on the basis of Payout and then using domain inputs find the relation between payout bucket and adjuster.
i am new to kind of domain specific problems and i would welcome all the input, suggestions ,anything from this community. As this is a very important project for me and client.