Year: 2023
Type: conference-paper
Abstract: Categorical features are a challenge for most machine learning algorithms that only accept numerical vectors in input.However, the emergence of graph neural networks is revolutionizing the application of machine learning models to traditional data sets.This is thanks to the possibility of introducing graph relationships amongst features and samples.In this contribution, we describe an algorithm leveraging the assignment matrix of a DiffPool graph neural network to calculate embeddings for categorical features, using as an adjacency matrix the co-occurrence matrix between the categorical values and as nodes feature the one hot encoded categorical values.We show that the algorithm proposed is scalable and presents a competitive performance in three publicly available data sets presenting both numerical and categorical values. 41
Authors: Weiwei Wang, Stefano Bromuri, Michel Dumontier
Institutions: Maastricht University, Open University of the Netherlands
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4386815495