Optimized Vehicle Repair Cost by Means of Smart Repair Distribution Model

The distribution of vehicle damages to bodyshops is not cost-effective. Every damage is unique, and every bodyshop has its specialization(s). Currently, the distribution of damages to bodyshops is predominately based on the distance between the customer and the bodyshop. This paper provides a method to optimize the distribution of vehicle damages to bodyshops in a way that the repair is executed by the most cost-effective bodyshop available for a particular damage based on damage and context characteristics. Three machine learning models have been evaluated to determine which is most suitable for predicting the cost of repair for one particular damage for each bodyshop available. The neural network produced the best results with an average error of €383. In order to apply this approach to real-world problems, we highlight the use of data from visual assessment of the damages using computer vision technology and onboard vehicle data in order to yield the biggest improvement in the average prediction error.