Exploring the Relationship Between Feature Attribution Methods and Model Performance

Author(s)

Silva, Priscylla ; Silva, Claudio ; Nonato, Luis Gustavo

• Document type: Journal article
• Source: NEURIPS WORKSHOPS, 2020; v. 257, p. 10-pg., 2024-01-01.

Abstract

Machine learning and deep learning models are pivotal in educational contexts, particularly in predicting student success. Despite their widespread application, a significant gap persists in comprehending the factors influencing these models' predictions, especially in explainability within education. This work addresses this gap by employing nine distinct explanation methods and conducting a comprehensive analysis to explore the correlation between the agreement among these methods in generating explanations and the predictive model's performance. Applying Spearman's correlation, our findings reveal a very strong correlation between the model's performance and the agreement level observed among the explanation methods. (AU)

• FAPESP's process: 23/05783-5 - Investigating the disagreement problem in local explanation methods
• Grantee: Priscylla Maria da Silva Sousa
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• FAPESP's process: 22/09091-8 - Criminality, insecurity, and legitimacy: a transdisciplinary approach
• Grantee: Luis Gustavo Nonato
• Support Opportunities: Research Grants - eScience and Data Science Program - Thematic Grants
• FAPESP's process: 22/03941-0 - An interpretable predictive model for crime forecasting using Graph Neural Network
• Grantee: Priscylla Maria da Silva Sousa
• Support Opportunities: Scholarships in Brazil - Doctorate