Learning to rank based on multi-criteria optimization
S. V. Semenikhin, Liudmila A. Denisova · 2017
In recent years, there has been growing interest in learning to rank. We considered the current state of learning to rank in information retrieval systems. We proposed an approach for learning to rank problem based on multi-criteria optimization using the method of Pareto optimization and Genetic Algorithms. The performance of the method has been investigated on test data collections, also a comparison with existing methods of learning to rank has been performed. Weight vector for a ranking function obtained in this research shows the effectiveness of the proposed method. Also we consider a possible direction of the further improvements of the learning to rank method proposed in this article.