hadoop

Hadoop Use Case: Auto Insurance Pricing Based on Driver Behavior

10 Jan 2015

The Situation

Insurance companies around the world strive to provide lower insurance rates, and auto insurance is no exception to this phenomenon. The automobile insurance companies are devising ways to derive innovative pricing models that will help customers reduce their insurance premiums; however, it requires an understanding of how one drives their vehicle. Insurance companies determine the premiums not just by the driver’s or vehicle’s history but also by statistical probabilities including age and gender.

Pay How You Drive (PHYD)

Usage-based insurance (UBI) also known as pay as you drive (PAYD) and pay how you drive (PHYD) and mile-based auto insurance are types of vehicle insurance whereby the costs are dependent upon type of vehicle used, measured against time, distance, behavior and place. This differs from traditional insurance, which attempts to differentiate and reward “safe” drivers by giving lower premiums and/or a no-claims bonus. PAYD car insurance enables the insurance companies to charge the client based on how they actually drive their vehicle. This radically changes the insurance model where distance driven and insurance premiums are not aligned.

This requires integration of several technologies to enable an end-to-end system including an in-vehicle device, communication network and an application platform hosting a number of capabilities including a Geographical Information System (GIS) and Customer Relationship Management (CRM) system. The European and North American motor insurance industries have been operating at a loss due to price pressure and the customer acquisitions costs related to the need to hold a statistically viable data set that actuaries can analyze to define risk against. PHYD based insurance is able to offer an alternative to this method by charging an insurance premium based on the actual risk of the driver and reduce the size and volume of claims by between 17% and 50%.

Implementation

PHYD requires an insurance company to source the driving data and is usually achieved by offering a traditional policy at a discount and by installing a telematics device. The driver’s data can be analyzed against existing risk measurement tools and insurance segmentation. Products can then be structured around charging for high-risk behaviors or reducing the premium for low risk. The complexities of the charging regime are only limited by the ability to communicate the proposition to the end user.

Ingenious Qube did an analysis on the data provided by its client, procured from the sensors installed on the vehicles. The sensors send the data every 5 seconds to the backend system, and then the client provided us the data files for further processing and data analysis. The analysis was based on user behavior on these aspects:

  • Over speed
  • Hard Breaking
  • Hard Acceleration

graph

The data has provided our client the ability to improve its services and to provide the best premium rates to its customers based on the above data.

This helps in defining actual risk rather than projected risk to a group of users based on their driving behavior. The transition from projected to actual risk is based on the insurer’s ability to identify new risk factors from the data and apply these within a product.

Driver activity data can be mapped and used to identify generic high-risk behaviors. This method of detecting risk by applying algorithms to automatically analyze the data is sometimes referred to as the driver fingerprinting or DNA.

insurance

Technology

Ingenious Qube helped the client in identifying which technology to be used for the proof of concept (POC). The implementation of the POC included the Hortonworks Data Platform (HDP) and specifically leveraging Apache HBase for NoSQL, Apache Pig scripting language, Apache Hive for SQL queries and Apache SQOOP for data transfer into Hadoop. Apache Solr was used for the Hadoop search capabilities.

The selection of Hortonworks Data Platform (HDP) over other distributions was primarily due to its open source nature and its various deployment options including cloud or an on-premise solution across both LINUX and Windows operating systems. HDP is also a complete Hadoop solution and incorporates all the components to create an enterprise platform including data access, data management, security, operations and governance.

Data received from the vehicle sensors was imported into HDP using SQOOP, an open source tool for importing structured data like the CSV files received from the client. Queries were then written using Hive and Pig to extract the required data based on predefined specific criteria of user driving behavior.

Apache Solr, an adjunct open source search tool to HDP, was used to search through the large chunks of data (Terabytes/Petabytes). The combined solution was able to process the large data volumes in a much faster manner as compared to traditional databases and provided many valuable insights.

Potential Benefits

  • Social and environmental benefits from more responsible driving
  • Commercial benefits to the insurance company from better alignment of insurance with actual risk. Improved customer segmentation
  • Potential cost-savings for responsible customers
  • Technology that powers UBI/PHYD enables other vehicle-to-infrastructure solutions including drive-through payments, emergency road assistance etc.
  • More choice for consumers on type of car insurance available to buy

About Hortonworks

Hortonworks is the only 100-percent open source software provider to develop, distribute and support an Apache Hadoop platform explicitly architected, built and tested for enterprise-grade deployments. Developed by the original architects, builders and operators of Hadoop, Hortonworks stewards the core and delivers the critical services required by the enterprise to reliably and effectively run Hadoop at scale. Our distribution, Hortonworks Data Platform, provides an open and stable foundation for enterprises and a growing ecosystem to build and deploy big data solutions. Hortonworks also provides unmatched technical support, training and certification programs. For more information and to try it out at no cost, download the Hortonworks Sandbox at http://hortonworks.com/sandbox.


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>