The use of machine learning technologies is optimal for this task since insurance decisions are largely driven by the previous history of applications and are based on massive data sets. Cognitive systems can help handlers review cases efficiently, evaluate cases with greater accuracy, and make more informed decisions.

The use of cognitive technologies not only increases the speed of data processing and the process of making informed decisions but also makes it possible to identify such claims at the stage of preliminary information analysis.

When processing an application, one of the most important tasks of the underwriter is risk assessment. The analysis of big data and the use of complex ML underwriting algorithms that determine implicit processes and patterns certainly exceed the capabilities of “classical” methods based only on statistical models.

One of the most relevant practical problems is the prediction of high-cost claimants a relatively small group of patients who account for a disproportionately large share of insurance claims. High medical costs for these patients often occur as part of emergency treatment, while there are options for earlier intervention to reduce costs and prevent acute conditions. The timely provision of such services is a win-win, both for the insurer who avoids increased costs and for the insured who receives quality medical care. The potential of predictive models based on ML methods to solve the problem of identifying such patients has been studied in several serious scientific publications.

Researchers and modellers agree that the predictive power of risk assessment and identification of patients with specific prospectively high-cost needs is cost-effective. The conclusions of the scientific and analytical community are successfully confirmed in practice.

Prediction accuracy increases due to an increase in the number of factors that affect the probability of an insured event, which, in turn, determines the cost of the contract. In addition, the big data accumulated during the interpretation of the model results also made it possible to identify new features and patterns, which became an additional advantage of the company’s product.