Spatiotemporal prediction 3-graph transformer

Mission

Predict future wind power

data

Each sample includes seq_x As input ,seq_y As label

1. Reading data csv
2. Before processing ,raw_df, new_df Increased time 、 Information about the day of the week , Removed tmstamp、day Etc
3. Intercept seq_x and seq_y, With new_df Benchmarking , according to index selection history and predict
4. establish graph

chart

adopt np.corrcoef and np.where Composition

np.corrcoef Calculate the correlation of each node ,134 * 134
np.argpartition Calculate the most relevant 5 A corresponding index

Model

Model combination graph and autoformer

Reference resources

• https://github.com/PaddlePaddle/PGL/tree/main/examples/kddcup2022/wpf_baseline
• https://keras.io/examples/graph/gat_node_classification/