Data Science in Insurance Domain
Since last decade, many business infrastructures are focusing on data collection and extracting useful information from data- the realm of data science. With transition from manual analysis towards and automation and availability of computational power, there is a rapid growth in business applications of data science.
One such scenario comes from domain of Insurance Analytics. In India, insurance industry is expected to reach USD$ 280 billion by 2020 (Source: https://www.ibef.org). However, insurance frauds have been rapidly increased. There are certain areas, where frauds claims have been so frequent that many insurance companies have blacklisted these areas. Also, cost incurred in manually checking each claims reliability is huge.
With availability of huge amount of data, focus has now shifted to identify fraud insurance claims with help of data science techniques. But how does it contribute to business ?
Ideally, solution would be to build best predictive model from the data provided. But, it would be more convenient to work around probability scores. Which means, focus is to build model which will give probabilities whether the claim is likely to be fraud or not. Also, we will bucket top 20% of scores and would tell client to investigate only these top 20% samples . Why so ? because in these top 20% chances of getting non-fraudulent claims are pretty much less and hence cost incurred in manually investigating is reduced, which leads to profitability.
Hence, when it comes to business, focus is not always on building accurate models, but improving profitability and reducing losses.
By: Gaurav Chavan