Kolloquium: Prediction of Pareto Fronts for Multimodal Travel Itineraries

Abstract:
Integrated mobility platforms promise travelers to create door-to-door itineraries considering their individual preferences considering the breadth of mobility services such as trains, buses, flights, and ridesharing services. Finding the complete Pareto-optimal set of itineraries with multiple traveler preferences in a multimodal setting is a significant challenge. We approximate the set of Pareto-optimal itineraries by solutionsampling equally distributed over the solution space. We also investigate offline learning (with a random-forest prediction model) and online learning (with Gaussian Process Regression) to make our solution framework smarter. Based on a large real-world data set, we investigate the performance and the effectiveness of our framework.