Insurance Price Prediction

Finance & Insurance

Background

The client has an insurance comparison service. A single comparison request engages a few dozens of conversations with web services for insurance companies, which takes time and is repeatedly unreliable. Slow web services cause delays, leading to a reduced quality of the service and lower conversion rates. The client needed to reduce the comparison response time and make it stable.

Our work

We designed a model, as an intermediate layer, to predict a price given by a web service with the input data (a few dozens of features, including high cardinality sets, such as a postal code or a car model).

The outcome

The prediction service we delivered is used as a supplementary infrastructure, which eliminates downtimes in the customer care process and increases the conversion rate.

Technology used
  • Scala
  • R
  • Python
  • TensorFlow