# Graph Node Classification
In this competition, you need to predict the classes of every node. There are 45,236 nodes and 4,585,028 links for training data, and 11,708 nodes and 359,180 links for testing data. 
We also provide graph id, which indicated the group of every nodes (training data contains 20 groups and testing data contains 4 groups). 
The dimension of features for nodes is 50.
There are 121 classes, and each node can be one-class, multi-class or non-class.

## Dataset Description
There are 4 files for training:
1. **train_graph.json**: graph information, i.e. 'nodes', 'links'
2. **train_graph_id.npy**: the index  for indicating the group of nodes (1-20)
3. **train_feats.npy**: the features for every nodes, dimension is 50.
4. **train_labels.npy**: the classes for every nodes, dimension is 121.

There are 4 files for training:
1. **test_graph.json**: graph information, i.e. 'nodes', 'links'
2. **test_graph_id.npy**: the index  for indicating the group of nodes (21-24)
3. **test_feats.npy**: the features for every nodes, dimension is 50.
Acknowledgements
Citation would be undated after the competition.