From User Activity Traces to Navigation Graph for Software Enhancement: An Application of Graph Neural Network (GNN) on a Real-World Non-Attributed Graph

What is it about?

User traces can be modeled in the form of graphs which reflect the user's navigation within software. Analyzing these graphs to elicit user behaviors requires the use of traditional machine learning methods. However, the navigation graphs generated from sets of traces are not suitable for the direct application of these methods (incomplete, unlabeled graphs, ... etc). To solve this problem, we opted for the application of Graph Neural Networks (GNNs) and the generation of artificial features of the nodes, the results obtained are promising and have proven the effectiveness of these features on a node classification task. The node classification task itself had the aim of understanding the composition of the pages/nodes constituting the software (always with the aim of improving software performance).

Featured Image

Photo by Markus Spiske on Unsplash

Photo by Markus Spiske on Unsplash