Network Information Propagation Prediction Model Based on Graph Attention Network with Transformer

In recent days, defined prediction of information propagation in networks playing a crucial role in the context of precision marketing. Traditional machine models are utilized for the prediction of information propagation but the accuracy of models depends on the designed features. However, the existing transformer based prediction model is implemented for information propagation but the features are handled limitedly. To overcome these issues, Graph Attention Network with Transformer (GAT-Trans) is proposed for the information propagation prediction. The Graph Attention Network (GAT) handles the features by updating the node representation and captures the time-evolving relationships. Primarily, the data is taken from Sina Weibo dataset and then preprocessed with one-hot encoding technique. Then, the contextualized features are extracted with transformer to improve the performance of prediction model. Finally, the prediction is done with the help of GAT model to predict the network information propagation into correctly propagated or affected for the future user nodes. From the results, the proposed GAT-Trans prediction model gave better results when compared to existing GCIP-Page Rank model by providing results such as accuracy (99.26%), precision (97.18%), recall (97.48%) and F1-score (98.37%) respectively.