Network embedding aims to map nodes in a network to low-dimensional vector representations. Graph neural networks (GNNs) have received much attention and have achieved state-of-the-art performance in learning node representation. Using fundamental sociological theories (status theory and balance theory) to model signed networks, basing GNN on learning node embedding has become a hot topic in signed network embedding. However, most GNNs fail to use edge weight information in signed networks, and most models cannot be directly used in weighted signed networks. We propose a novel signed directed graph neural networks model named WSNN to learn node embedding for Weighted signed networks. The proposed model reconstructs link signs, link directions, and signed directed triangles simultaneously. Based on the network representations learned by the proposed model, we conduct link sign prediction in signed networks. Extensive experimental results in real-world datasets demonstrate the superiority of the proposed model over the state-of-the-art network embedding algorithms for graph representation learning in signed networks.

This paper proposes a feature selection model based on a multilayer genetic algorithm (GA) to select the features of a high stock dividend (HSD) and eliminate the relatively redundant features in the optimal solution by using layer-by-layer information transfer and two-dimensionality reduction methods. Combining the ensemble model and time-series split cross-validation (TSCV) indicator as the fitness function solves the problem of selecting the fitness function for each layer. The symmetry character of the model is fully utilized in the two-dimensionality reduction processes, according to the change in data dimensions and the unbalanced characteristics of the HSD, setting the corresponding TSCV indicators. We built seven ensemble prediction models for actual stock trading data for comparison experiments. The results show that the feature selection model based on multilayer GA can effectively eliminate the relatively redundant features after dimensionality reduction and significantly improve the balancing accuracy, precision and AUC performance of the seven ensemble learning models. Finally, adversarial validation is used to analyze the differences in the balanced accuracy of the training and test sets caused by the inconsistent distribution of the data sets.