Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
167 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

GC-LSTM: Graph Convolution Embedded LSTM for Dynamic Link Prediction (1812.04206v2)

Published 11 Dec 2018 in cs.SI and physics.soc-ph

Abstract: Dynamic link prediction is a research hot in complex networks area, especially for its wide applications in biology, social network, economy and industry. Compared with static link prediction, dynamic one is much more difficult since network structure evolves over time. Currently most researches focus on static link prediction which cannot achieve expected performance in dynamic network. Aiming at low AUC, high Error Rate, add/remove link prediction difficulty, we propose GC-LSTM, a Graph Convolution Network (GC) embedded Long Short Term Memory network (LTSM), for end-to-end dynamic link prediction. To the best of our knowledge, it is the first time that GCN embedded LSTM is put forward for link prediction of dynamic networks. GCN in this new deep model is capable of node structure learning of network snapshot for each time slide, while LSTM is responsible for temporal feature learning for network snapshot. Besides, current dynamic link prediction method can only handle removed links, GC-LSTM can predict both added or removed link at the same time. Extensive experiments are carried out to testify its performance in aspects of prediction accuracy, Error Rate, add/remove link prediction and key link prediction. The results prove that GC-LSTM outperforms current state-of-art method.

Citations (183)

Summary

  • The paper presents GC-LSTM, an end-to-end model that integrates GCN and LSTM to predict both link additions and deletions in dynamic networks.
  • It combines spatial feature extraction via graph convolution with temporal dynamics captured by LSTM, outperforming baseline methods on real datasets.
  • The model achieves high accuracy on metrics like AUC and Error Rate, demonstrating its potential for applications in social, economic, and biological network analysis.

GC-LSTM: Graph Convolution Embedded LSTM for Dynamic Network Link Prediction

The paper "GC-LSTM: Graph Convolution Embedded LSTM for Dynamic Network Link Prediction" presents a novel approach to dynamic network link prediction by integrating Graph Convolution Networks (GCN) with Long Short-Term Memory (LSTM) networks to form an end-to-end model named GC-LSTM. This integration is aimed at overcoming the limitations of traditional methods by effectively capturing both spatial and temporal features of dynamic networks.

Methodology and Model Architecture

GC-LSTM is specifically designed to predict changes in links within dynamic networks, a task that involves forecasting both the addition and removal of edges over time. Traditional models largely focus on static networks or are limited to predicting only one type of link change (e.g., additions). The novelty of GC-LSTM lies in its ability to handle both link additions and deletions by embedding GCN into LSTM cells. This modular approach allows the model to learn structural features through GCN and capture temporal dynamics via LSTM.

The model uses graph convolution to embed spatial relationships of each snapshot into LSTM, thereby enhancing its ability to learn from high-dimensional and time-dependent data. Important components of the GC-LSTM architecture include:

  • Graph Convolution: Applied to each network snapshot to capture local structural properties and node relationships.
  • LSTM: Used as the primary framework to encode temporal dependencies in the graph sequence over time.

Experimental Results and Evaluation

The paper reports extensive experiments conducted on six real-world dynamic network datasets, such as ENRON and LKML. Across these datasets, the GC-LSTM model significantly outperforms several baseline methods including node2vec, TNE, and DDNE in various metrics such as AUC, GMAUC, and Error Rate (ER).

A comprehensive evaluation reveals that GC-LSTM achieves superior performance in terms of accuracy, especially in predicting link evolutions that involve significant structural changes. The model's ability to effectively decrease prediction error rates while capturing both newly formed and dissolved links underscores its robustness compared to other state-of-the-art models.

Implications and Future Prospects

The integration of GCN with LSTM in the GC-LSTM model exemplifies a promising direction in dynamic network analysis. This method not only advances prediction accuracy but also contributes to a deeper understanding of dynamic interactions in networks such as social, economic, and biological systems. Its implications for real-world applications are vast, ranging from predicting social connections and customer relationships to monitoring disease spread through dynamic contact networks.

Future work can optimize GC-LSTM for large-scale dynamic networks involving a greater number of nodes and links. Additionally, tuning hyperparameters further and exploring variations of LSTM and GCN architectures may refine the model's efficiency and predictive power.

In conclusion, GC-LSTM provides a compelling framework for dynamic network analysis by bridging spatial and temporal relational modeling, thereby laying a groundwork for improved forecasting in complex and evolving systems.