Success Story: Deep Code: Deep Learning system for customer characterisation, prediction and decision making in the insurance industry

NCC presenting the success story

NCC Spain

Client/user profile:

A company which is part of a group technology companies with in-depth knowledge of the insurance sector. They offer their clients (insurers and brokers) the technological platforms and solutions that lead to a significant improvement in their productivity

Impact:

This system allows the company to offer insurers a technological platform to position themselves in the market and visualise their current situation, detect premium variations in the market, perform portfolio qualification through premium prediction or estimation systems, anticipate market variations.

Key points before agreeing on project:

• Data confidentiality
• IPR
• Efficient implementation
• Secured infrastructure access
• Migration to client infrastructure

Technical/scientific Challenge:

The challenge is to develop a Deep Learning system for the insurance sector to improve the data analysis needed in the actuarial process.

Insurance companies use complex statistical and mathematical models to estimate the different types of risk associated with an insurance policy and therefore to determine the possible premiums, with a whole science behind it, actuarial science.

In the case of automobile insurance, the risk estimation models can include from the variables most directly associated with the insured risk itself (characteristics of the vehicle, age and experience of the driver) to any other information that the company has directly or indirectly, such as the probability of non-payment or the number of kilometers that the insured is typically going to drive.

There is also the possibility of using statistical information, indirectly linked to the customer, but which has a direct impact on risk.

Solution:

The models to be developed must work with a large volume of data, as more than 200 million records are generated annually, so it is necessary to work with multiprocessor systems, either High Performance Computing (HPC) or Cloud Computing.

Deep Learning models for premium estimation, price trends and alerts have been developed to form a system that will provide the insurance sector with an analytical environment to support decision-making. The use of a Cloud Computing environment allowed the training and efficient execution of the models.

Benefits:

Make decisions and strategic changes based on data.