In Data-Transformation-Code folder (the folder where the dataset is located)
- Unzip the zip files
- Run gen_socialadj_listsv2.py
- Run format_tweetersv3_compressed.py
Choose to either:
A) Put all training data into history lists: (This gives all interations as history to the model)
3)Run format_retweetersv3.py
4)Run format_negsamplesv3.py (Create negative retweeter examples and format appropriately)
B) Split training data 1/4 goint into history lists, 3/4 held out: (This splits some interations for history other for future interactions) Temporal makes sure that history is consistent with time stamps
3)Run hist_format_retweetersv3.py or format_retweeters_temporal.py(latest)
4)Run hist_format_negsamplesv3.py or format_negsamplesv3_temporal.py(latest) (Create negative retweeter examples and format appropriately)
- Run gen_temporal_history_timeupdate.py (This orders interactions by time, making sure the interaction events are in temporal order)
6)Run format_noninteractersv2.py (Just need to do this so index errors aren't thrown)
-
Copy gendata folder over to GraphRec-WWW19 folder
-
Run either Debug_GraphRec.py or Tune_GraphRec.py
Important Variables: pos_to_neg_interaction_dict maps each positive interaction to its associated sampled negative interactions for use during training