Ponente
Descripción
There is an increased demand for tools capable of helping users to access relevant information rapidly. Companies have a high interest in data-based recommendations to maximize client satisfaction. The need to make it easier to use machine learning to solve all kinds of data-based tasks has led to a boom in techniques for automating this process. The present study proposes using an automated machine learning system (AutoML) to solve ranking problems through the AutoGOAL library. All the functionalities are necessary to solve basic tasks of what is known as Learning to Rank (LTR) are adapted to this library. These new features are metrics, input types, and algorithms explicitly dedicated to LTR tasks. After a few iterations in the hyperparameter optimization process, the proposal presents positive results on the data sets worked on. The values obtained are favourable compared to the evaluations of these algorithms outside the AutoGOAL system. An implementation of AutoGOAL with the ranking algorithm library is freely available on GitHub.
Keywords: AutoGOAL, Recommender Systems, Automated Machine Learning, Learning to Rank.
