KDD Cup 2017

Project Organization

• bnlstm: An implementation of Recurrent Batch Normalization in TensorFlow
• config: Model config and dataset description
• data.checkpoint: Tensorflow model files
• data.dataSets: Rawdata, in csv format
• data.features: Temporal data, in Pickle format
• data.prediction: Prediction results
• dataloader: Data iterator
• feature: Functions concern feature preprocessing
• model: Machine learning models
• util: I/O and other utility functions

Requirements

• Tensorflow 1.0
• python 3.5

Get started

• Download dataset, unpack and move them to data.dataSets directory
• Check and edit all the fields related to dataset and experiment settings in config.py

References

1. Kipf T N, Welling M. Semi-supervised classification with graph convolutional networks[J]. arXiv preprint arXiv:1609.02907, 2016.
2. Cooijmans T, Ballas N, Laurent C, et al. Recurrent batch normalization[J]. arXiv preprint arXiv:1603.09025, 2016.
3. Shahsavari B, Abbeel P. Short-term traffic forecasting: Modeling and learning spatio-temporal relations in transportation networks using graph neural networks[J]. 2015.
4. Della Valle E, Celino I, Dell’Aglio D, et al. Urban Computing: a challenging problem for Semantic Technologies[C]//2nd International Workshop on New Forms of Reasoning for the Semantic Web (NEFORS 2008) co-located with the 3rd Asian Semantic Web Conference (ASWC 2008). 2008.
5. Che Z, Purushotham S, Cho K, et al. Recurrent neural networks for multivariate time series with missing values[J]. arXiv preprint arXiv:1606.01865, 2016.