An Immune-Inspired Technique to Identify Heavy Goods Vehicles Incident Hot Spots

We report on the adaptation of an immune-inspired instance selection technique to solve a real-world big data problem of determining vehicle incident hot spots. The technique, which is inspired by the immune system self-regulation mechanism, was originally conceptualized to eliminate very similar instances in data classification tasks. We adapt the method to detect hot spots from a telematics dataset containing hundreds of thousands of data points indicating incident locations involving heavy goods vehicles (HGVs) across the United Kingdom. The objective is to provide HGV drivers with information regarding areas of high likelihood of incidents in order to continuously improve road safety and vehicle economy. The problem presents several challenges and constraints. An accurate view of the hot spots produced in a timely manner is necessary. In addition, the solution is required to be adaptable and dynamic, as thousands of new incidents are included in the database daily. Furthermore, the impact on hot spots after informing drivers about their existence has to be considered. Our solution successfully addresses these constraints. It is fast, robust, and applicable to all different incidents investigated. The method is also self-adjustable, which means that if the hot spots configuration changes with time, the algorithm automatically evolves to present the most current topology. Our solution has been implemented by a telematics company to improve HGV safety in the United Kingdom.