Prescriptive Analytics on VRP through Reinforcement Learning and Event Streams

Abstract

Business processes require quick decisions to constantly adapt to changes in order to improve performance and take advantage of opportunities. It is essential to have analytics that transform data into knowledge for decision making. This paper introduces a line of research focused on prescriptive analytics, capable of calculating actions to be executed at the moment (operational decisions) or in the future (tactical and/or strategic decisions) to achieve a desired goal, in vehicle routing problems (VRP), and presents the progress and results obtained. The calculation of the actions involves the processing of the flow of business events in the form of datastreams, the application of Soft Computing and Computational Intelligence techniques and algorithms (in particular Reinforcement Learning) and, derived from the need for low response times, the use of High Performance Computing.