Applications of machine learning techniques on graph data

Summary

In recent years advancements in Machine Learning and the massive production of data with the natural occurrence of relations lead organizations and research to seek new methods to exploit them or generalize traditional machine learning technologies for this purpose. The most well known occurrence is social networks. The purpose of this thesis is to present and explain the basic principles and terms related to the domain of Graph Representation Learning.

We present all basic terms needed for the reader to understand all the algorithms explained in the literature and explain the inner workings of those algorithms according to characteristics and taxonomies. In the use case of fraud detection and the learning objective of semi supervised learning we choose the most appropriate algorithms based on the characteristics that are useful for this scenario and we assess their performance based on the selected use case and the dataset chosen. We perform and comment on the results of several experiments designed to test the performance of those algorithms of the GRL domain in those specific data.