This is the code for submission #8692 in AAAI2020
The geolife data to evaluate our model, which contains 40 users and ready for directly used.
- Python 3.6
- TF-1.18
- /data # preprocessed
- pos.train.txt
- pos.validate.txt
- pos.test.txt
- pos.vocab.txt
- region_data
- region_distance_no_less10.npy
- region_beijing_grid2region_REID_no_less10.json
- /model # AttnMove
- main.py
- dataset_utils.py
- Embedder.py
- model_utils.py
- self_attn_hyperaram_region.py
- /baseline #codes for baselines
- baseline.py
- dataset_utils.py
- Embedder.py
- model_utils.py
- self_attn_hyperaram_region.py
- /log #save dir
- analyse_data # save attention map & location embedding
python3 main.py --blank_num 10 --if_history 1 --hidden_dim 64 --bleu_interval 5 --nhead 1 --nlayer 2 --reg_lambda 1e-2 --gpu 4 --if_mask 1 --drop 0.3 --fb_drop 0.3
The codes contain four network model (AttnMove, AttnMove-H) and baseline model (LSTM, Bi-LSTM, DeepMove)