A Multi-Objective Learning to re-Rank Approach to Optimize Online\n Marketplaces for Multiple Stakeholders

Multi-objective recommender systems address the difficult task of\nrecommending items that are relevant to multiple, possibly conflicting,\ncriteria. However these systems are most often designed to address the\nobjective of one single stakeholder, typically, in online commerce, the\nconsumers whose input and purchasing decisions ultimately determine the success\nof the recommendation systems. In this work, we address the multi-objective,\nmulti-stakeholder, recommendation problem involving one or more objective(s)\nper stakeholder. In addition to the consumer stakeholder, we also consider two\nother stakeholders; the suppliers who provide the goods and services for sale\nand the intermediary who is responsible for helping connect consumers to\nsuppliers via its recommendation algorithms. We analyze the multi-objective,\nmulti-stakeholder, problem from the point of view of the online marketplace\nintermediary whose objective is to maximize its commission through its\nrecommender system. We define a multi-objective problem relating all our three\nstakeholders which we solve with a novel learning-to-re-rank approach that\nmakes use of a novel regularization function based on the Kendall tau\ncorrelation metric and its kernel version; given an initial ranking of item\nrecommendations built for the consumer, we aim to re-rank it such that the new\nranking is also optimized for the secondary objectives while staying close to\nthe initial ranking. We evaluate our approach on a real-world dataset of hotel\nrecommendations provided by Expedia where we show the effectiveness of our\napproach against a business-rules oriented baseline model.\n