Research Article
Relationship Discovery and Hierarchical Embedding for Web Service Quality Prediction
Table 1
Default main hyper parameters setting.
| Modules | Main hyper parameters | Value |
| LDA | Number of topics in LDA | 20 | Dirichlet parameter | 0.1 |
| Node2vec | Length of Node2vec walk | 80 | Number of Node2vec walks | 10 | Window size of Node2vec walk | 10 |
| GCN | Number of layers | 2 | Size of layer convolution layer | 64 |
| Wide & deep | Number of layers | 2 | Dimensions | 128 |
| GraphSAGE | Neighborhood deep neighborhood sample sizes | 2 | 25 |
| NGCF | Embedding propagation layers | 3 | Depth of the NGCF | 3 | Top-k | 20 |
| PinSage | Top-k | 2 | Size of hidden dimension size | 512 |
| KGAT | Top-k | 20 | Tower structure of hidden layer size | 512/256/128/64 |
| GraphRec | Size of the hidden layer | 128 | Number of three hidden layers | 3 |
| NIA-GCN | Size of first layer/second layer | 40/2 | Dimension of each layer | 64 |
|
|