On the explainability of Graph Neural Networks: an application in credit scoring
Title | On the explainability of Graph Neural Networks: an application in credit scoring |
---|---|
Summary | On the explainability of Graph Neural Networks: an application in credit scoring |
Keywords | Graph Neural Networks, Explainable AI, Credit Scoring |
TimeFrame | Fall 2023 |
References | [[References::[1] Feng, Bojing, et al. "Every Corporation Owns Its Structure: Corporate Credit Rating via Graph Neural Networks." Chinese Conference on Pattern Recognition and Computer Vision (PRCV). Cham: Springer International Publishing, 2022.
[2] http://web.stanford.edu/class/cs224w/index.html#content [3] Yuan, Hao, et al. "Explainability in graph neural networks: A taxonomic survey." IEEE transactions on pattern analysis and machine intelligence 45.5 (2022): 5782-5799.]] |
Prerequisites | |
Author | |
Supervisor | Atiye Sadat Hashemi, Peyman Mashhadi |
Level | Master |
Status | Open |
Graph Neural Networks (GNNs) lies in the intersection of graph theory and neural networks, enabling domain knowledge to be incorporated into the design of a model through graph connectivity [2]. The encoded domain knowledge into the graph structure gives another dimension that neutral networks can build upon. The black box nature of deep neural networks accompanied with the additional encoded information in forms of graph give rise to more complexity. XAI methods [3] can help unwrap this black box nature. In the financial sector, credit scoring (a numerical expression based on a level analysis of a person's credit files, to represent the creditworthiness of an individual) is an important application which providers of these services demand explanability from AI methods [1]. In this project, besides GNNs, XAI techniques specialized for GNNs will be explored.