Graph Neural networks

AI and machine learning are commonly seen as areas of great interest because of their promise in improving business results and creating new impact. Graph can be used to augment data science in a few key ways.

Most commonly, features for machine learning can be created via graph by running graph algorithms on a dataset that has been loaded into a graph database, and creating enriched data which can then be used for machine learning. This step of feature engineering provides the machine learning model with more comprehensive, useful information.

Many data scientists are starting to become interested in graph neural networks, which can capture the graph itself as an input of machine learning and neural networks. The graph can potentially hold more information than standard tables because of the flexibility of the model. Machine learning models with information captured from graphs often provide better performance than machine learning based on table shape input.