GITD: Enhancing Medical Classification on Tabular Data with Missing Values via Graph Modeling
Keywords: medical classification, tabular data, imputation
TL;DR: We propose a novel approach that enables the use of graph data imputation methods on medical tabular data.
Abstract: With the advancement of machine learning, various techniques have been developed to classify patients for disease diagnosis using medical tabular data. Due to the presence of missing values in the medical tabular data, these techniques commonly impute the missing values before applying classifiers. However, most existing techniques classify patients solely based on each patient's individual features despite the advantages of leveraging patients with similar features that can enhance both imputation and classification. To address this issue, we introduce graph data imputation for tabular data (GITD), a novel approach that constructs feature-attentive k-nearest neighbor (kNN) graphs to enable the use of graph data imputation methods on medical tabular data. The key idea of GITD is constructing a kNN graph among patients by prioritizing important features for classification. Our extensive experimental results demonstrate that GITD successfully bridges graph data imputation methods and medical tabular classification, achieving state-of-the-art performance across various medical tabular datasets.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7618
Loading