Our clients wanted to have one or more algorithms built that could use large, healthcare paid claim data sets and identify those claims most likely to be part of an accident, as well as most likely to NOT be part of an accident. It is possible that there may need to be separate algorithms for the distinct accident types. They had the claims data and their outcomes data.
We developed an algorithm, which could be inserted into their existing ETL process, and provided a single, key metric: likelihood of this claim being part of an accident. We demonstrated the expected claims whose decisions changed between actual historical decision and modeled decision so that we could estimate profitability lift and ensure a sense of comfort in the expected outcome of the model.