Structural-topic aware deep neural networks for information cascade prediction.

Area of Science:

• Social Network Analysis
• Information Science
• Machine Learning

Background:

• Accurate prediction of information cascade popularity is crucial for applications like online opinion analysis.
• Existing deep learning models often overlook the structural information within cascade networks.
• Bridging the gap between prediction accuracy and interpretability in cascade modeling remains a challenge.

Purpose of the Study:

• To propose a novel deep neural network model that integrates structural information for enhanced cascade prediction.
• To leverage both structural and topical features for more accurate forecasting of information cascade spread.
• To develop a model that combines the interpretability of traditional methods with the predictive power of deep learning.

Main Methods:

• Developed Structural-Topic Aware Deep Neural Networks (STDNN) to capture node structure and topic distributions.
• Employed a sequential neural network to process learned structural-topic features for prediction.
• Integrated graph structure analysis with deep learning techniques for cascade popularity forecasting.

Main Results:

• STDNN demonstrated promising performance in predicting future information cascade popularity.
• The proposed model achieved higher efficiency compared to existing baseline methods.
• Quantitative experiments validated the effectiveness of incorporating structural information.

Conclusions:

• STDNN successfully integrates network structure and topic distributions for improved cascade prediction.
• The model offers a balance between high predictive power and interpretability.
• This approach advances the understanding and prediction of information cascades by utilizing graph structures.