Optimization of Control Agents Shifts in Public Transportation: Tackling Fare Evasion with Machine-Learning

In this article, we present a research project aiming at tackling fare evasion in public transportation by optimizing the action of control agents. We give an overview of an algorithm that combines reinforcement learning techniques with optimization methods in order to predict which are the areas of the network where fraud is particularly high and generate itineraries accordingly. The proposed solution combines public and private data and is intended to be suited for most transportation operators worldwide. Its first deployment territory will be in the region of Paris (2018).