CC-GCN: enhancing graph convolutional networks using neighborhood interactions

Graph neural networks (GNNs) are a class of deep learning approaches designed for graph based data inference that have made noteworthy strides in handling graphical datasets in past few years. Notably, graph convolutional networks (GCNs) have demonstrated strong performance even with limited training data, often surpassing traditional state-of-the-art models across various tasks, including semi-supervised node classification, link prediction, graph classification, as well as applications in natural language processing and computational chemistry. In this work, we suggest sampling the training set nodes based on their clustering tendency rather than using a conventional random sample selection method. Then, we input the sampled nodes into GCN to carry out semi-supervised node classification tasks. Additionally, we provide a mathematical justification for how our suggested sample selection method leads to the creation of improved GNN models. This leads to better training and consequently better node embeddings and better outputs. Based on this unique training technique, the performance of GCN model over three benchmark networks—Cora, Citeseer, and PubMed has been examined. Following the implementation of the GCN model, classification accuracies of $$90.12\%$$ for CORA, $$89.20\%$$ for PubMed, and $$77.0\%$$ for Citeseer were achieved. CC-GCN attains the best accuracy among the evaluated methods on CORA and PubMed and remains competitive on Citeseer, indicating the effectiveness of structurally informed training-node selection. We also acquired an accuracy of $$91.85\%$$ on Amazon Computers dataset and $$84.04\%$$ on WikiCS dataset. We have also detailed how GCN based multimedia evaluation works for social networks and we have proved this by applying it to hostility classification of Hindi tweets. We have obtained weighted F1-score of 0.64, which is competitive with reported systems for this task. Our findings are generic and task-independent. Additionally, the suggested method can yield positive outcomes for other tasks, such as community detection and connection prediction.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-12
DOI
https://doi.org/10.1038/s41598-026-74980-9
Primary Topic
Advanced Graph Neural Networks
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

CC-GCN: enhancing graph convolutional networks using neighborhood interactions

Mahipal Jadeja, Rahul Saxena, Atul Kumar Verma
Scientific Reports
Advanced Graph Neural Networks
article

CC-GCN: enhancing graph convolutional networks using neighborhood interactions

Mahipal Jadeja, Rahul Saxena, Atul Kumar Verma
article en

Abstract

Graph neural networks (GNNs) are a class of deep learning approaches designed for graph based data inference that have made noteworthy strides in handling graphical datasets in past few years. Notably, graph convolutional networks (GCNs) have demonstrated strong performance even with limited training data, often surpassing traditional state-of-the-art models across various tasks, including semi-supervised node classification, link prediction, graph classification, as well as applications in natural language processing and computational chemistry. In this work, we suggest sampling the training set nodes based on their clustering tendency rather than using a conventional random sample selection method. Then, we input the sampled nodes into GCN to carry out semi-supervised node classification tasks. Additionally, we provide a mathematical justification for how our suggested sample selection method leads to the creation of improved GNN models. This leads to better training and consequently better node embeddings and better outputs. Based on this unique training technique, the performance of GCN model over three benchmark networks—Cora, Citeseer, and PubMed has been examined. Following the implementation of the GCN model, classification accuracies of $$90.12\%$$ for CORA, $$89.20\%$$ for PubMed, and $$77.0\%$$ for Citeseer were achieved. CC-GCN attains the best accuracy among the evaluated methods on CORA and PubMed and remains competitive on Citeseer, indicating the effectiveness of structurally informed training-node selection. We also acquired an accuracy of $$91.85\%$$ on Amazon Computers dataset and $$84.04\%$$ on WikiCS dataset. We have also detailed how GCN based multimedia evaluation works for social networks and we have proved this by applying it to hostility classification of Hindi tweets. We have obtained weighted F1-score of 0.64, which is competitive with reported systems for this task. Our findings are generic and task-independent. Additionally, the suggested method can yield positive outcomes for other tasks, such as community detection and connection prediction.

Scientific Reports
Manipal University Jaipur (IN), Malaviya National Institute of Technology Jaipur (IN)
Advanced Graph Neural Networks
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon