FSD-CAP: Fractional Subgraph Diffusion with Class-Aware Propagation for Graph Feature Imputation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Qiao, Xin, Sun, Shijie, Dong, Anqi, Hua, Cong, Zhao, Xia, Zhang, Longfei, Zhu, Guangming, Zhang, Liang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915757425688576
author Qiao, Xin
Sun, Shijie
Dong, Anqi
Hua, Cong
Zhao, Xia
Zhang, Longfei
Zhu, Guangming
Zhang, Liang
author_facet Qiao, Xin
Sun, Shijie
Dong, Anqi
Hua, Cong
Zhao, Xia
Zhang, Longfei
Zhu, Guangming
Zhang, Liang
contents Imputing missing node features in graphs is challenging, particularly under high missing rates. Existing methods based on latent representations or global diffusion often fail to produce reliable estimates, and may propagate errors across the graph. We propose FSD-CAP, a two-stage framework designed to improve imputation quality under extreme sparsity. In the first stage, a graph-distance-guided subgraph expansion localizes the diffusion process. A fractional diffusion operator adjusts propagation sharpness based on local structure. In the second stage, imputed features are refined using class-aware propagation, which incorporates pseudo-labels and neighborhood entropy to promote consistency. We evaluated FSD-CAP on multiple datasets. With $99.5\%$ of features missing across five benchmark datasets, FSD-CAP achieves average accuracies of $80.06\%$ (structural) and $81.01\%$ (uniform) in node classification, close to the $81.31\%$ achieved by a standard GCN with full features. For link prediction under the same setting, it reaches AUC scores of $91.65\%$ (structural) and $92.41\%$ (uniform), compared to $95.06\%$ for the fully observed case. Furthermore, FSD-CAP demonstrates superior performance on both large-scale and heterophily datasets when compared to other models.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18938
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FSD-CAP: Fractional Subgraph Diffusion with Class-Aware Propagation for Graph Feature Imputation
Qiao, Xin
Sun, Shijie
Dong, Anqi
Hua, Cong
Zhao, Xia
Zhang, Longfei
Zhu, Guangming
Zhang, Liang
Machine Learning
Social and Information Networks
68T07, 68Q32, 90C26
Imputing missing node features in graphs is challenging, particularly under high missing rates. Existing methods based on latent representations or global diffusion often fail to produce reliable estimates, and may propagate errors across the graph. We propose FSD-CAP, a two-stage framework designed to improve imputation quality under extreme sparsity. In the first stage, a graph-distance-guided subgraph expansion localizes the diffusion process. A fractional diffusion operator adjusts propagation sharpness based on local structure. In the second stage, imputed features are refined using class-aware propagation, which incorporates pseudo-labels and neighborhood entropy to promote consistency. We evaluated FSD-CAP on multiple datasets. With $99.5\%$ of features missing across five benchmark datasets, FSD-CAP achieves average accuracies of $80.06\%$ (structural) and $81.01\%$ (uniform) in node classification, close to the $81.31\%$ achieved by a standard GCN with full features. For link prediction under the same setting, it reaches AUC scores of $91.65\%$ (structural) and $92.41\%$ (uniform), compared to $95.06\%$ for the fully observed case. Furthermore, FSD-CAP demonstrates superior performance on both large-scale and heterophily datasets when compared to other models.
title FSD-CAP: Fractional Subgraph Diffusion with Class-Aware Propagation for Graph Feature Imputation
topic Machine Learning
Social and Information Networks
68T07, 68Q32, 90C26
url https://arxiv.org/abs/2601.18938